xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 7259ca31048e5ced8e7f90657a3d7084aeafdf51)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002,2003 Henning Brauer
6  * Copyright (c) 2012 Gleb Smirnoff <glebius@FreeBSD.org>
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  *	$OpenBSD: pf_ioctl.c,v 1.213 2009/02/15 21:46:12 mbalmer Exp $
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_inet.h"
44 #include "opt_inet6.h"
45 #include "opt_bpf.h"
46 #include "opt_pf.h"
47 
48 #include <sys/param.h>
49 #include <sys/_bitset.h>
50 #include <sys/bitset.h>
51 #include <sys/bus.h>
52 #include <sys/conf.h>
53 #include <sys/endian.h>
54 #include <sys/fcntl.h>
55 #include <sys/filio.h>
56 #include <sys/hash.h>
57 #include <sys/interrupt.h>
58 #include <sys/jail.h>
59 #include <sys/kernel.h>
60 #include <sys/kthread.h>
61 #include <sys/lock.h>
62 #include <sys/mbuf.h>
63 #include <sys/module.h>
64 #include <sys/nv.h>
65 #include <sys/proc.h>
66 #include <sys/sdt.h>
67 #include <sys/smp.h>
68 #include <sys/socket.h>
69 #include <sys/sysctl.h>
70 #include <sys/md5.h>
71 #include <sys/ucred.h>
72 
73 #include <net/if.h>
74 #include <net/if_var.h>
75 #include <net/vnet.h>
76 #include <net/route.h>
77 #include <net/pfil.h>
78 #include <net/pfvar.h>
79 #include <net/if_pfsync.h>
80 #include <net/if_pflog.h>
81 
82 #include <netinet/in.h>
83 #include <netinet/ip.h>
84 #include <netinet/ip_var.h>
85 #include <netinet6/ip6_var.h>
86 #include <netinet/ip_icmp.h>
87 #include <netpfil/pf/pf_nv.h>
88 
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #endif /* INET6 */
92 
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96 
97 SDT_PROBE_DEFINE3(pf, ioctl, ioctl, error, "int", "int", "int");
98 SDT_PROBE_DEFINE3(pf, ioctl, function, error, "char *", "int", "int");
99 SDT_PROBE_DEFINE2(pf, ioctl, addrule, error, "int", "int");
100 SDT_PROBE_DEFINE2(pf, ioctl, nvchk, error, "int", "int");
101 
102 static struct pf_kpool	*pf_get_kpool(const char *, u_int32_t, u_int8_t,
103 			    u_int32_t, u_int8_t, u_int8_t, u_int8_t);
104 
105 static void		 pf_mv_kpool(struct pf_kpalist *, struct pf_kpalist *);
106 static void		 pf_empty_kpool(struct pf_kpalist *);
107 static int		 pfioctl(struct cdev *, u_long, caddr_t, int,
108 			    struct thread *);
109 #ifdef ALTQ
110 static int		 pf_begin_altq(u_int32_t *);
111 static int		 pf_rollback_altq(u_int32_t);
112 static int		 pf_commit_altq(u_int32_t);
113 static int		 pf_enable_altq(struct pf_altq *);
114 static int		 pf_disable_altq(struct pf_altq *);
115 static uint16_t		 pf_qname2qid(const char *);
116 static void		 pf_qid_unref(uint16_t);
117 #endif /* ALTQ */
118 static int		 pf_begin_rules(u_int32_t *, int, const char *);
119 static int		 pf_rollback_rules(u_int32_t, int, char *);
120 static int		 pf_setup_pfsync_matching(struct pf_kruleset *);
121 static void		 pf_hash_rule(MD5_CTX *, struct pf_krule *);
122 static void		 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
123 static int		 pf_commit_rules(u_int32_t, int, char *);
124 static int		 pf_addr_setup(struct pf_kruleset *,
125 			    struct pf_addr_wrap *, sa_family_t);
126 static void		 pf_addr_copyout(struct pf_addr_wrap *);
127 static void		 pf_src_node_copy(const struct pf_ksrc_node *,
128 			    struct pf_src_node *);
129 #ifdef ALTQ
130 static int		 pf_export_kaltq(struct pf_altq *,
131 			    struct pfioc_altq_v1 *, size_t);
132 static int		 pf_import_kaltq(struct pfioc_altq_v1 *,
133 			    struct pf_altq *, size_t);
134 #endif /* ALTQ */
135 
136 VNET_DEFINE(struct pf_krule,	pf_default_rule);
137 
138 #ifdef ALTQ
139 VNET_DEFINE_STATIC(int,		pf_altq_running);
140 #define	V_pf_altq_running	VNET(pf_altq_running)
141 #endif
142 
143 #define	TAGID_MAX	 50000
144 struct pf_tagname {
145 	TAILQ_ENTRY(pf_tagname)	namehash_entries;
146 	TAILQ_ENTRY(pf_tagname)	taghash_entries;
147 	char			name[PF_TAG_NAME_SIZE];
148 	uint16_t		tag;
149 	int			ref;
150 };
151 
152 struct pf_tagset {
153 	TAILQ_HEAD(, pf_tagname)	*namehash;
154 	TAILQ_HEAD(, pf_tagname)	*taghash;
155 	unsigned int			 mask;
156 	uint32_t			 seed;
157 	BITSET_DEFINE(, TAGID_MAX)	 avail;
158 };
159 
160 VNET_DEFINE(struct pf_tagset, pf_tags);
161 #define	V_pf_tags	VNET(pf_tags)
162 static unsigned int	pf_rule_tag_hashsize;
163 #define	PF_RULE_TAG_HASH_SIZE_DEFAULT	128
164 SYSCTL_UINT(_net_pf, OID_AUTO, rule_tag_hashsize, CTLFLAG_RDTUN,
165     &pf_rule_tag_hashsize, PF_RULE_TAG_HASH_SIZE_DEFAULT,
166     "Size of pf(4) rule tag hashtable");
167 
168 #ifdef ALTQ
169 VNET_DEFINE(struct pf_tagset, pf_qids);
170 #define	V_pf_qids	VNET(pf_qids)
171 static unsigned int	pf_queue_tag_hashsize;
172 #define	PF_QUEUE_TAG_HASH_SIZE_DEFAULT	128
173 SYSCTL_UINT(_net_pf, OID_AUTO, queue_tag_hashsize, CTLFLAG_RDTUN,
174     &pf_queue_tag_hashsize, PF_QUEUE_TAG_HASH_SIZE_DEFAULT,
175     "Size of pf(4) queue tag hashtable");
176 #endif
177 VNET_DEFINE(uma_zone_t,	 pf_tag_z);
178 #define	V_pf_tag_z		 VNET(pf_tag_z)
179 static MALLOC_DEFINE(M_PFALTQ, "pf_altq", "pf(4) altq configuration db");
180 static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
181 
182 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
183 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
184 #endif
185 
186 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
187 			    unsigned int);
188 static void		 pf_cleanup_tagset(struct pf_tagset *);
189 static uint16_t		 tagname2hashindex(const struct pf_tagset *, const char *);
190 static uint16_t		 tag2hashindex(const struct pf_tagset *, uint16_t);
191 static u_int16_t	 tagname2tag(struct pf_tagset *, const char *);
192 static u_int16_t	 pf_tagname2tag(const char *);
193 static void		 tag_unref(struct pf_tagset *, u_int16_t);
194 
195 #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x
196 
197 struct cdev *pf_dev;
198 
199 /*
200  * XXX - These are new and need to be checked when moveing to a new version
201  */
202 static void		 pf_clear_all_states(void);
203 static unsigned int	 pf_clear_states(const struct pf_kstate_kill *);
204 static void		 pf_killstates(struct pf_kstate_kill *,
205 			    unsigned int *);
206 static int		 pf_killstates_row(struct pf_kstate_kill *,
207 			    struct pf_idhash *);
208 static int		 pf_killstates_nv(struct pfioc_nv *);
209 static int		 pf_clearstates_nv(struct pfioc_nv *);
210 static int		 pf_getstate(struct pfioc_nv *);
211 static int		 pf_getstatus(struct pfioc_nv *);
212 static int		 pf_clear_tables(void);
213 static void		 pf_clear_srcnodes(struct pf_ksrc_node *);
214 static void		 pf_kill_srcnodes(struct pfioc_src_node_kill *);
215 static int		 pf_keepcounters(struct pfioc_nv *);
216 static void		 pf_tbladdr_copyout(struct pf_addr_wrap *);
217 
218 /*
219  * Wrapper functions for pfil(9) hooks
220  */
221 #ifdef INET
222 static pfil_return_t pf_check_in(struct mbuf **m, struct ifnet *ifp,
223     int flags, void *ruleset __unused, struct inpcb *inp);
224 static pfil_return_t pf_check_out(struct mbuf **m, struct ifnet *ifp,
225     int flags, void *ruleset __unused, struct inpcb *inp);
226 #endif
227 #ifdef INET6
228 static pfil_return_t pf_check6_in(struct mbuf **m, struct ifnet *ifp,
229     int flags, void *ruleset __unused, struct inpcb *inp);
230 static pfil_return_t pf_check6_out(struct mbuf **m, struct ifnet *ifp,
231     int flags, void *ruleset __unused, struct inpcb *inp);
232 #endif
233 
234 static void		hook_pf(void);
235 static void		dehook_pf(void);
236 static int		shutdown_pf(void);
237 static int		pf_load(void);
238 static void		pf_unload(void);
239 
240 static struct cdevsw pf_cdevsw = {
241 	.d_ioctl =	pfioctl,
242 	.d_name =	PF_NAME,
243 	.d_version =	D_VERSION,
244 };
245 
246 volatile VNET_DEFINE_STATIC(int, pf_pfil_hooked);
247 #define V_pf_pfil_hooked	VNET(pf_pfil_hooked)
248 
249 /*
250  * We need a flag that is neither hooked nor running to know when
251  * the VNET is "valid".  We primarily need this to control (global)
252  * external event, e.g., eventhandlers.
253  */
254 VNET_DEFINE(int, pf_vnet_active);
255 #define V_pf_vnet_active	VNET(pf_vnet_active)
256 
257 int pf_end_threads;
258 struct proc *pf_purge_proc;
259 
260 struct rmlock			pf_rules_lock;
261 struct sx			pf_ioctl_lock;
262 struct sx			pf_end_lock;
263 
264 /* pfsync */
265 VNET_DEFINE(pfsync_state_import_t *, pfsync_state_import_ptr);
266 VNET_DEFINE(pfsync_insert_state_t *, pfsync_insert_state_ptr);
267 VNET_DEFINE(pfsync_update_state_t *, pfsync_update_state_ptr);
268 VNET_DEFINE(pfsync_delete_state_t *, pfsync_delete_state_ptr);
269 VNET_DEFINE(pfsync_clear_states_t *, pfsync_clear_states_ptr);
270 VNET_DEFINE(pfsync_defer_t *, pfsync_defer_ptr);
271 pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr;
272 
273 /* pflog */
274 pflog_packet_t			*pflog_packet_ptr = NULL;
275 
276 /*
277  * Copy a user-provided string, returning an error if truncation would occur.
278  * Avoid scanning past "sz" bytes in the source string since there's no
279  * guarantee that it's nul-terminated.
280  */
281 static int
282 pf_user_strcpy(char *dst, const char *src, size_t sz)
283 {
284 	if (strnlen(src, sz) == sz)
285 		return (EINVAL);
286 	(void)strlcpy(dst, src, sz);
287 	return (0);
288 }
289 
290 static void
291 pfattach_vnet(void)
292 {
293 	u_int32_t *my_timeout = V_pf_default_rule.timeout;
294 
295 	pf_initialize();
296 	pfr_initialize();
297 	pfi_initialize_vnet();
298 	pf_normalize_init();
299 	pf_syncookies_init();
300 
301 	V_pf_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
302 	V_pf_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
303 
304 	RB_INIT(&V_pf_anchors);
305 	pf_init_kruleset(&pf_main_ruleset);
306 
307 	/* default rule should never be garbage collected */
308 	V_pf_default_rule.entries.tqe_prev = &V_pf_default_rule.entries.tqe_next;
309 #ifdef PF_DEFAULT_TO_DROP
310 	V_pf_default_rule.action = PF_DROP;
311 #else
312 	V_pf_default_rule.action = PF_PASS;
313 #endif
314 	V_pf_default_rule.nr = -1;
315 	V_pf_default_rule.rtableid = -1;
316 
317 	pf_counter_u64_init(&V_pf_default_rule.evaluations, M_WAITOK);
318 	for (int i = 0; i < 2; i++) {
319 		pf_counter_u64_init(&V_pf_default_rule.packets[i], M_WAITOK);
320 		pf_counter_u64_init(&V_pf_default_rule.bytes[i], M_WAITOK);
321 	}
322 	V_pf_default_rule.states_cur = counter_u64_alloc(M_WAITOK);
323 	V_pf_default_rule.states_tot = counter_u64_alloc(M_WAITOK);
324 	V_pf_default_rule.src_nodes = counter_u64_alloc(M_WAITOK);
325 
326 #ifdef PF_WANT_32_TO_64_COUNTER
327 	V_pf_kifmarker = malloc(sizeof(*V_pf_kifmarker), PFI_MTYPE, M_WAITOK | M_ZERO);
328 	V_pf_rulemarker = malloc(sizeof(*V_pf_rulemarker), M_PFRULE, M_WAITOK | M_ZERO);
329 	PF_RULES_WLOCK();
330 	LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist);
331 	LIST_INSERT_HEAD(&V_pf_allrulelist, &V_pf_default_rule, allrulelist);
332 	V_pf_allrulecount++;
333 	LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist);
334 	PF_RULES_WUNLOCK();
335 #endif
336 
337 	/* initialize default timeouts */
338 	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
339 	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
340 	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
341 	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
342 	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
343 	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
344 	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
345 	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
346 	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
347 	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
348 	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
349 	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
350 	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
351 	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
352 	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
353 	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
354 	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
355 	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
356 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
357 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
358 
359 	bzero(&V_pf_status, sizeof(V_pf_status));
360 	V_pf_status.debug = PF_DEBUG_URGENT;
361 
362 	V_pf_pfil_hooked = 0;
363 
364 	/* XXX do our best to avoid a conflict */
365 	V_pf_status.hostid = arc4random();
366 
367 	for (int i = 0; i < PFRES_MAX; i++)
368 		V_pf_status.counters[i] = counter_u64_alloc(M_WAITOK);
369 	for (int i = 0; i < KLCNT_MAX; i++)
370 		V_pf_status.lcounters[i] = counter_u64_alloc(M_WAITOK);
371 	for (int i = 0; i < FCNT_MAX; i++)
372 		pf_counter_u64_init(&V_pf_status.fcounters[i], M_WAITOK);
373 	for (int i = 0; i < SCNT_MAX; i++)
374 		V_pf_status.scounters[i] = counter_u64_alloc(M_WAITOK);
375 
376 	if (swi_add(&V_pf_swi_ie, "pf send", pf_intr, curvnet, SWI_NET,
377 	    INTR_MPSAFE, &V_pf_swi_cookie) != 0)
378 		/* XXXGL: leaked all above. */
379 		return;
380 }
381 
382 static struct pf_kpool *
383 pf_get_kpool(const char *anchor, u_int32_t ticket, u_int8_t rule_action,
384     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
385     u_int8_t check_ticket)
386 {
387 	struct pf_kruleset	*ruleset;
388 	struct pf_krule		*rule;
389 	int			 rs_num;
390 
391 	ruleset = pf_find_kruleset(anchor);
392 	if (ruleset == NULL)
393 		return (NULL);
394 	rs_num = pf_get_ruleset_number(rule_action);
395 	if (rs_num >= PF_RULESET_MAX)
396 		return (NULL);
397 	if (active) {
398 		if (check_ticket && ticket !=
399 		    ruleset->rules[rs_num].active.ticket)
400 			return (NULL);
401 		if (r_last)
402 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
403 			    pf_krulequeue);
404 		else
405 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
406 	} else {
407 		if (check_ticket && ticket !=
408 		    ruleset->rules[rs_num].inactive.ticket)
409 			return (NULL);
410 		if (r_last)
411 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
412 			    pf_krulequeue);
413 		else
414 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
415 	}
416 	if (!r_last) {
417 		while ((rule != NULL) && (rule->nr != rule_number))
418 			rule = TAILQ_NEXT(rule, entries);
419 	}
420 	if (rule == NULL)
421 		return (NULL);
422 
423 	return (&rule->rpool);
424 }
425 
426 static void
427 pf_mv_kpool(struct pf_kpalist *poola, struct pf_kpalist *poolb)
428 {
429 	struct pf_kpooladdr	*mv_pool_pa;
430 
431 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
432 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
433 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
434 	}
435 }
436 
437 static void
438 pf_empty_kpool(struct pf_kpalist *poola)
439 {
440 	struct pf_kpooladdr *pa;
441 
442 	while ((pa = TAILQ_FIRST(poola)) != NULL) {
443 		switch (pa->addr.type) {
444 		case PF_ADDR_DYNIFTL:
445 			pfi_dynaddr_remove(pa->addr.p.dyn);
446 			break;
447 		case PF_ADDR_TABLE:
448 			/* XXX: this could be unfinished pooladdr on pabuf */
449 			if (pa->addr.p.tbl != NULL)
450 				pfr_detach_table(pa->addr.p.tbl);
451 			break;
452 		}
453 		if (pa->kif)
454 			pfi_kkif_unref(pa->kif);
455 		TAILQ_REMOVE(poola, pa, entries);
456 		free(pa, M_PFRULE);
457 	}
458 }
459 
460 static void
461 pf_unlink_rule(struct pf_krulequeue *rulequeue, struct pf_krule *rule)
462 {
463 
464 	PF_RULES_WASSERT();
465 
466 	TAILQ_REMOVE(rulequeue, rule, entries);
467 
468 	PF_UNLNKDRULES_LOCK();
469 	rule->rule_ref |= PFRULE_REFS;
470 	TAILQ_INSERT_TAIL(&V_pf_unlinked_rules, rule, entries);
471 	PF_UNLNKDRULES_UNLOCK();
472 }
473 
474 void
475 pf_free_rule(struct pf_krule *rule)
476 {
477 
478 	PF_RULES_WASSERT();
479 
480 	if (rule->tag)
481 		tag_unref(&V_pf_tags, rule->tag);
482 	if (rule->match_tag)
483 		tag_unref(&V_pf_tags, rule->match_tag);
484 #ifdef ALTQ
485 	if (rule->pqid != rule->qid)
486 		pf_qid_unref(rule->pqid);
487 	pf_qid_unref(rule->qid);
488 #endif
489 	switch (rule->src.addr.type) {
490 	case PF_ADDR_DYNIFTL:
491 		pfi_dynaddr_remove(rule->src.addr.p.dyn);
492 		break;
493 	case PF_ADDR_TABLE:
494 		pfr_detach_table(rule->src.addr.p.tbl);
495 		break;
496 	}
497 	switch (rule->dst.addr.type) {
498 	case PF_ADDR_DYNIFTL:
499 		pfi_dynaddr_remove(rule->dst.addr.p.dyn);
500 		break;
501 	case PF_ADDR_TABLE:
502 		pfr_detach_table(rule->dst.addr.p.tbl);
503 		break;
504 	}
505 	if (rule->overload_tbl)
506 		pfr_detach_table(rule->overload_tbl);
507 	if (rule->kif)
508 		pfi_kkif_unref(rule->kif);
509 	pf_kanchor_remove(rule);
510 	pf_empty_kpool(&rule->rpool.list);
511 
512 	pf_krule_free(rule);
513 }
514 
515 static void
516 pf_init_tagset(struct pf_tagset *ts, unsigned int *tunable_size,
517     unsigned int default_size)
518 {
519 	unsigned int i;
520 	unsigned int hashsize;
521 
522 	if (*tunable_size == 0 || !powerof2(*tunable_size))
523 		*tunable_size = default_size;
524 
525 	hashsize = *tunable_size;
526 	ts->namehash = mallocarray(hashsize, sizeof(*ts->namehash), M_PFHASH,
527 	    M_WAITOK);
528 	ts->taghash = mallocarray(hashsize, sizeof(*ts->taghash), M_PFHASH,
529 	    M_WAITOK);
530 	ts->mask = hashsize - 1;
531 	ts->seed = arc4random();
532 	for (i = 0; i < hashsize; i++) {
533 		TAILQ_INIT(&ts->namehash[i]);
534 		TAILQ_INIT(&ts->taghash[i]);
535 	}
536 	BIT_FILL(TAGID_MAX, &ts->avail);
537 }
538 
539 static void
540 pf_cleanup_tagset(struct pf_tagset *ts)
541 {
542 	unsigned int i;
543 	unsigned int hashsize;
544 	struct pf_tagname *t, *tmp;
545 
546 	/*
547 	 * Only need to clean up one of the hashes as each tag is hashed
548 	 * into each table.
549 	 */
550 	hashsize = ts->mask + 1;
551 	for (i = 0; i < hashsize; i++)
552 		TAILQ_FOREACH_SAFE(t, &ts->namehash[i], namehash_entries, tmp)
553 			uma_zfree(V_pf_tag_z, t);
554 
555 	free(ts->namehash, M_PFHASH);
556 	free(ts->taghash, M_PFHASH);
557 }
558 
559 static uint16_t
560 tagname2hashindex(const struct pf_tagset *ts, const char *tagname)
561 {
562 	size_t len;
563 
564 	len = strnlen(tagname, PF_TAG_NAME_SIZE - 1);
565 	return (murmur3_32_hash(tagname, len, ts->seed) & ts->mask);
566 }
567 
568 static uint16_t
569 tag2hashindex(const struct pf_tagset *ts, uint16_t tag)
570 {
571 
572 	return (tag & ts->mask);
573 }
574 
575 static u_int16_t
576 tagname2tag(struct pf_tagset *ts, const char *tagname)
577 {
578 	struct pf_tagname	*tag;
579 	u_int32_t		 index;
580 	u_int16_t		 new_tagid;
581 
582 	PF_RULES_WASSERT();
583 
584 	index = tagname2hashindex(ts, tagname);
585 	TAILQ_FOREACH(tag, &ts->namehash[index], namehash_entries)
586 		if (strcmp(tagname, tag->name) == 0) {
587 			tag->ref++;
588 			return (tag->tag);
589 		}
590 
591 	/*
592 	 * new entry
593 	 *
594 	 * to avoid fragmentation, we do a linear search from the beginning
595 	 * and take the first free slot we find.
596 	 */
597 	new_tagid = BIT_FFS(TAGID_MAX, &ts->avail);
598 	/*
599 	 * Tags are 1-based, with valid tags in the range [1..TAGID_MAX].
600 	 * BIT_FFS() returns a 1-based bit number, with 0 indicating no bits
601 	 * set.  It may also return a bit number greater than TAGID_MAX due
602 	 * to rounding of the number of bits in the vector up to a multiple
603 	 * of the vector word size at declaration/allocation time.
604 	 */
605 	if ((new_tagid == 0) || (new_tagid > TAGID_MAX))
606 		return (0);
607 
608 	/* Mark the tag as in use.  Bits are 0-based for BIT_CLR() */
609 	BIT_CLR(TAGID_MAX, new_tagid - 1, &ts->avail);
610 
611 	/* allocate and fill new struct pf_tagname */
612 	tag = uma_zalloc(V_pf_tag_z, M_NOWAIT);
613 	if (tag == NULL)
614 		return (0);
615 	strlcpy(tag->name, tagname, sizeof(tag->name));
616 	tag->tag = new_tagid;
617 	tag->ref = 1;
618 
619 	/* Insert into namehash */
620 	TAILQ_INSERT_TAIL(&ts->namehash[index], tag, namehash_entries);
621 
622 	/* Insert into taghash */
623 	index = tag2hashindex(ts, new_tagid);
624 	TAILQ_INSERT_TAIL(&ts->taghash[index], tag, taghash_entries);
625 
626 	return (tag->tag);
627 }
628 
629 static void
630 tag_unref(struct pf_tagset *ts, u_int16_t tag)
631 {
632 	struct pf_tagname	*t;
633 	uint16_t		 index;
634 
635 	PF_RULES_WASSERT();
636 
637 	index = tag2hashindex(ts, tag);
638 	TAILQ_FOREACH(t, &ts->taghash[index], taghash_entries)
639 		if (tag == t->tag) {
640 			if (--t->ref == 0) {
641 				TAILQ_REMOVE(&ts->taghash[index], t,
642 				    taghash_entries);
643 				index = tagname2hashindex(ts, t->name);
644 				TAILQ_REMOVE(&ts->namehash[index], t,
645 				    namehash_entries);
646 				/* Bits are 0-based for BIT_SET() */
647 				BIT_SET(TAGID_MAX, tag - 1, &ts->avail);
648 				uma_zfree(V_pf_tag_z, t);
649 			}
650 			break;
651 		}
652 }
653 
654 static uint16_t
655 pf_tagname2tag(const char *tagname)
656 {
657 	return (tagname2tag(&V_pf_tags, tagname));
658 }
659 
660 #ifdef ALTQ
661 static uint16_t
662 pf_qname2qid(const char *qname)
663 {
664 	return (tagname2tag(&V_pf_qids, qname));
665 }
666 
667 static void
668 pf_qid_unref(uint16_t qid)
669 {
670 	tag_unref(&V_pf_qids, qid);
671 }
672 
673 static int
674 pf_begin_altq(u_int32_t *ticket)
675 {
676 	struct pf_altq	*altq, *tmp;
677 	int		 error = 0;
678 
679 	PF_RULES_WASSERT();
680 
681 	/* Purge the old altq lists */
682 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
683 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
684 			/* detach and destroy the discipline */
685 			error = altq_remove(altq);
686 		}
687 		free(altq, M_PFALTQ);
688 	}
689 	TAILQ_INIT(V_pf_altq_ifs_inactive);
690 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
691 		pf_qid_unref(altq->qid);
692 		free(altq, M_PFALTQ);
693 	}
694 	TAILQ_INIT(V_pf_altqs_inactive);
695 	if (error)
696 		return (error);
697 	*ticket = ++V_ticket_altqs_inactive;
698 	V_altqs_inactive_open = 1;
699 	return (0);
700 }
701 
702 static int
703 pf_rollback_altq(u_int32_t ticket)
704 {
705 	struct pf_altq	*altq, *tmp;
706 	int		 error = 0;
707 
708 	PF_RULES_WASSERT();
709 
710 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
711 		return (0);
712 	/* Purge the old altq lists */
713 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
714 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
715 			/* detach and destroy the discipline */
716 			error = altq_remove(altq);
717 		}
718 		free(altq, M_PFALTQ);
719 	}
720 	TAILQ_INIT(V_pf_altq_ifs_inactive);
721 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
722 		pf_qid_unref(altq->qid);
723 		free(altq, M_PFALTQ);
724 	}
725 	TAILQ_INIT(V_pf_altqs_inactive);
726 	V_altqs_inactive_open = 0;
727 	return (error);
728 }
729 
730 static int
731 pf_commit_altq(u_int32_t ticket)
732 {
733 	struct pf_altqqueue	*old_altqs, *old_altq_ifs;
734 	struct pf_altq		*altq, *tmp;
735 	int			 err, error = 0;
736 
737 	PF_RULES_WASSERT();
738 
739 	if (!V_altqs_inactive_open || ticket != V_ticket_altqs_inactive)
740 		return (EBUSY);
741 
742 	/* swap altqs, keep the old. */
743 	old_altqs = V_pf_altqs_active;
744 	old_altq_ifs = V_pf_altq_ifs_active;
745 	V_pf_altqs_active = V_pf_altqs_inactive;
746 	V_pf_altq_ifs_active = V_pf_altq_ifs_inactive;
747 	V_pf_altqs_inactive = old_altqs;
748 	V_pf_altq_ifs_inactive = old_altq_ifs;
749 	V_ticket_altqs_active = V_ticket_altqs_inactive;
750 
751 	/* Attach new disciplines */
752 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
753 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
754 			/* attach the discipline */
755 			error = altq_pfattach(altq);
756 			if (error == 0 && V_pf_altq_running)
757 				error = pf_enable_altq(altq);
758 			if (error != 0)
759 				return (error);
760 		}
761 	}
762 
763 	/* Purge the old altq lists */
764 	TAILQ_FOREACH_SAFE(altq, V_pf_altq_ifs_inactive, entries, tmp) {
765 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
766 			/* detach and destroy the discipline */
767 			if (V_pf_altq_running)
768 				error = pf_disable_altq(altq);
769 			err = altq_pfdetach(altq);
770 			if (err != 0 && error == 0)
771 				error = err;
772 			err = altq_remove(altq);
773 			if (err != 0 && error == 0)
774 				error = err;
775 		}
776 		free(altq, M_PFALTQ);
777 	}
778 	TAILQ_INIT(V_pf_altq_ifs_inactive);
779 	TAILQ_FOREACH_SAFE(altq, V_pf_altqs_inactive, entries, tmp) {
780 		pf_qid_unref(altq->qid);
781 		free(altq, M_PFALTQ);
782 	}
783 	TAILQ_INIT(V_pf_altqs_inactive);
784 
785 	V_altqs_inactive_open = 0;
786 	return (error);
787 }
788 
789 static int
790 pf_enable_altq(struct pf_altq *altq)
791 {
792 	struct ifnet		*ifp;
793 	struct tb_profile	 tb;
794 	int			 error = 0;
795 
796 	if ((ifp = ifunit(altq->ifname)) == NULL)
797 		return (EINVAL);
798 
799 	if (ifp->if_snd.altq_type != ALTQT_NONE)
800 		error = altq_enable(&ifp->if_snd);
801 
802 	/* set tokenbucket regulator */
803 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
804 		tb.rate = altq->ifbandwidth;
805 		tb.depth = altq->tbrsize;
806 		error = tbr_set(&ifp->if_snd, &tb);
807 	}
808 
809 	return (error);
810 }
811 
812 static int
813 pf_disable_altq(struct pf_altq *altq)
814 {
815 	struct ifnet		*ifp;
816 	struct tb_profile	 tb;
817 	int			 error;
818 
819 	if ((ifp = ifunit(altq->ifname)) == NULL)
820 		return (EINVAL);
821 
822 	/*
823 	 * when the discipline is no longer referenced, it was overridden
824 	 * by a new one.  if so, just return.
825 	 */
826 	if (altq->altq_disc != ifp->if_snd.altq_disc)
827 		return (0);
828 
829 	error = altq_disable(&ifp->if_snd);
830 
831 	if (error == 0) {
832 		/* clear tokenbucket regulator */
833 		tb.rate = 0;
834 		error = tbr_set(&ifp->if_snd, &tb);
835 	}
836 
837 	return (error);
838 }
839 
840 static int
841 pf_altq_ifnet_event_add(struct ifnet *ifp, int remove, u_int32_t ticket,
842     struct pf_altq *altq)
843 {
844 	struct ifnet	*ifp1;
845 	int		 error = 0;
846 
847 	/* Deactivate the interface in question */
848 	altq->local_flags &= ~PFALTQ_FLAG_IF_REMOVED;
849 	if ((ifp1 = ifunit(altq->ifname)) == NULL ||
850 	    (remove && ifp1 == ifp)) {
851 		altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
852 	} else {
853 		error = altq_add(ifp1, altq);
854 
855 		if (ticket != V_ticket_altqs_inactive)
856 			error = EBUSY;
857 
858 		if (error)
859 			free(altq, M_PFALTQ);
860 	}
861 
862 	return (error);
863 }
864 
865 void
866 pf_altq_ifnet_event(struct ifnet *ifp, int remove)
867 {
868 	struct pf_altq	*a1, *a2, *a3;
869 	u_int32_t	 ticket;
870 	int		 error = 0;
871 
872 	/*
873 	 * No need to re-evaluate the configuration for events on interfaces
874 	 * that do not support ALTQ, as it's not possible for such
875 	 * interfaces to be part of the configuration.
876 	 */
877 	if (!ALTQ_IS_READY(&ifp->if_snd))
878 		return;
879 
880 	/* Interrupt userland queue modifications */
881 	if (V_altqs_inactive_open)
882 		pf_rollback_altq(V_ticket_altqs_inactive);
883 
884 	/* Start new altq ruleset */
885 	if (pf_begin_altq(&ticket))
886 		return;
887 
888 	/* Copy the current active set */
889 	TAILQ_FOREACH(a1, V_pf_altq_ifs_active, entries) {
890 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
891 		if (a2 == NULL) {
892 			error = ENOMEM;
893 			break;
894 		}
895 		bcopy(a1, a2, sizeof(struct pf_altq));
896 
897 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
898 		if (error)
899 			break;
900 
901 		TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, a2, entries);
902 	}
903 	if (error)
904 		goto out;
905 	TAILQ_FOREACH(a1, V_pf_altqs_active, entries) {
906 		a2 = malloc(sizeof(*a2), M_PFALTQ, M_NOWAIT);
907 		if (a2 == NULL) {
908 			error = ENOMEM;
909 			break;
910 		}
911 		bcopy(a1, a2, sizeof(struct pf_altq));
912 
913 		if ((a2->qid = pf_qname2qid(a2->qname)) == 0) {
914 			error = EBUSY;
915 			free(a2, M_PFALTQ);
916 			break;
917 		}
918 		a2->altq_disc = NULL;
919 		TAILQ_FOREACH(a3, V_pf_altq_ifs_inactive, entries) {
920 			if (strncmp(a3->ifname, a2->ifname,
921 				IFNAMSIZ) == 0) {
922 				a2->altq_disc = a3->altq_disc;
923 				break;
924 			}
925 		}
926 		error = pf_altq_ifnet_event_add(ifp, remove, ticket, a2);
927 		if (error)
928 			break;
929 
930 		TAILQ_INSERT_TAIL(V_pf_altqs_inactive, a2, entries);
931 	}
932 
933 out:
934 	if (error != 0)
935 		pf_rollback_altq(ticket);
936 	else
937 		pf_commit_altq(ticket);
938 }
939 #endif /* ALTQ */
940 
941 static int
942 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
943 {
944 	struct pf_kruleset	*rs;
945 	struct pf_krule		*rule;
946 
947 	PF_RULES_WASSERT();
948 
949 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
950 		return (EINVAL);
951 	rs = pf_find_or_create_kruleset(anchor);
952 	if (rs == NULL)
953 		return (EINVAL);
954 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
955 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
956 		rs->rules[rs_num].inactive.rcount--;
957 	}
958 	*ticket = ++rs->rules[rs_num].inactive.ticket;
959 	rs->rules[rs_num].inactive.open = 1;
960 	return (0);
961 }
962 
963 static int
964 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
965 {
966 	struct pf_kruleset	*rs;
967 	struct pf_krule		*rule;
968 
969 	PF_RULES_WASSERT();
970 
971 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
972 		return (EINVAL);
973 	rs = pf_find_kruleset(anchor);
974 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
975 	    rs->rules[rs_num].inactive.ticket != ticket)
976 		return (0);
977 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
978 		pf_unlink_rule(rs->rules[rs_num].inactive.ptr, rule);
979 		rs->rules[rs_num].inactive.rcount--;
980 	}
981 	rs->rules[rs_num].inactive.open = 0;
982 	return (0);
983 }
984 
985 #define PF_MD5_UPD(st, elm)						\
986 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
987 
988 #define PF_MD5_UPD_STR(st, elm)						\
989 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
990 
991 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
992 		(stor) = htonl((st)->elm);				\
993 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
994 } while (0)
995 
996 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
997 		(stor) = htons((st)->elm);				\
998 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
999 } while (0)
1000 
1001 static void
1002 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
1003 {
1004 	PF_MD5_UPD(pfr, addr.type);
1005 	switch (pfr->addr.type) {
1006 		case PF_ADDR_DYNIFTL:
1007 			PF_MD5_UPD(pfr, addr.v.ifname);
1008 			PF_MD5_UPD(pfr, addr.iflags);
1009 			break;
1010 		case PF_ADDR_TABLE:
1011 			PF_MD5_UPD(pfr, addr.v.tblname);
1012 			break;
1013 		case PF_ADDR_ADDRMASK:
1014 			/* XXX ignore af? */
1015 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
1016 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
1017 			break;
1018 	}
1019 
1020 	PF_MD5_UPD(pfr, port[0]);
1021 	PF_MD5_UPD(pfr, port[1]);
1022 	PF_MD5_UPD(pfr, neg);
1023 	PF_MD5_UPD(pfr, port_op);
1024 }
1025 
1026 static void
1027 pf_hash_rule(MD5_CTX *ctx, struct pf_krule *rule)
1028 {
1029 	u_int16_t x;
1030 	u_int32_t y;
1031 
1032 	pf_hash_rule_addr(ctx, &rule->src);
1033 	pf_hash_rule_addr(ctx, &rule->dst);
1034 	for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++)
1035 		PF_MD5_UPD_STR(rule, label[i]);
1036 	PF_MD5_UPD_STR(rule, ifname);
1037 	PF_MD5_UPD_STR(rule, match_tagname);
1038 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
1039 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
1040 	PF_MD5_UPD_HTONL(rule, prob, y);
1041 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
1042 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
1043 	PF_MD5_UPD(rule, uid.op);
1044 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
1045 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
1046 	PF_MD5_UPD(rule, gid.op);
1047 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
1048 	PF_MD5_UPD(rule, action);
1049 	PF_MD5_UPD(rule, direction);
1050 	PF_MD5_UPD(rule, af);
1051 	PF_MD5_UPD(rule, quick);
1052 	PF_MD5_UPD(rule, ifnot);
1053 	PF_MD5_UPD(rule, match_tag_not);
1054 	PF_MD5_UPD(rule, natpass);
1055 	PF_MD5_UPD(rule, keep_state);
1056 	PF_MD5_UPD(rule, proto);
1057 	PF_MD5_UPD(rule, type);
1058 	PF_MD5_UPD(rule, code);
1059 	PF_MD5_UPD(rule, flags);
1060 	PF_MD5_UPD(rule, flagset);
1061 	PF_MD5_UPD(rule, allow_opts);
1062 	PF_MD5_UPD(rule, rt);
1063 	PF_MD5_UPD(rule, tos);
1064 }
1065 
1066 static bool
1067 pf_krule_compare(struct pf_krule *a, struct pf_krule *b)
1068 {
1069 	MD5_CTX		ctx[2];
1070 	u_int8_t	digest[2][PF_MD5_DIGEST_LENGTH];
1071 
1072 	MD5Init(&ctx[0]);
1073 	MD5Init(&ctx[1]);
1074 	pf_hash_rule(&ctx[0], a);
1075 	pf_hash_rule(&ctx[1], b);
1076 	MD5Final(digest[0], &ctx[0]);
1077 	MD5Final(digest[1], &ctx[1]);
1078 
1079 	return (memcmp(digest[0], digest[1], PF_MD5_DIGEST_LENGTH) == 0);
1080 }
1081 
1082 static int
1083 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1084 {
1085 	struct pf_kruleset	*rs;
1086 	struct pf_krule		*rule, **old_array, *tail;
1087 	struct pf_krulequeue	*old_rules;
1088 	int			 error;
1089 	u_int32_t		 old_rcount;
1090 
1091 	PF_RULES_WASSERT();
1092 
1093 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1094 		return (EINVAL);
1095 	rs = pf_find_kruleset(anchor);
1096 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1097 	    ticket != rs->rules[rs_num].inactive.ticket)
1098 		return (EBUSY);
1099 
1100 	/* Calculate checksum for the main ruleset */
1101 	if (rs == &pf_main_ruleset) {
1102 		error = pf_setup_pfsync_matching(rs);
1103 		if (error != 0)
1104 			return (error);
1105 	}
1106 
1107 	/* Swap rules, keep the old. */
1108 	old_rules = rs->rules[rs_num].active.ptr;
1109 	old_rcount = rs->rules[rs_num].active.rcount;
1110 	old_array = rs->rules[rs_num].active.ptr_array;
1111 
1112 	rs->rules[rs_num].active.ptr =
1113 	    rs->rules[rs_num].inactive.ptr;
1114 	rs->rules[rs_num].active.ptr_array =
1115 	    rs->rules[rs_num].inactive.ptr_array;
1116 	rs->rules[rs_num].active.rcount =
1117 	    rs->rules[rs_num].inactive.rcount;
1118 
1119 	/* Attempt to preserve counter information. */
1120 	if (V_pf_status.keep_counters) {
1121 		TAILQ_FOREACH(rule, rs->rules[rs_num].active.ptr,
1122 		    entries) {
1123 			tail = TAILQ_FIRST(old_rules);
1124 			while ((tail != NULL) && ! pf_krule_compare(tail, rule))
1125 				tail = TAILQ_NEXT(tail, entries);
1126 			if (tail != NULL) {
1127 				pf_counter_u64_critical_enter();
1128 				pf_counter_u64_add_protected(&rule->evaluations,
1129 				    pf_counter_u64_fetch(&tail->evaluations));
1130 				pf_counter_u64_add_protected(&rule->packets[0],
1131 				    pf_counter_u64_fetch(&tail->packets[0]));
1132 				pf_counter_u64_add_protected(&rule->packets[1],
1133 				    pf_counter_u64_fetch(&tail->packets[1]));
1134 				pf_counter_u64_add_protected(&rule->bytes[0],
1135 				    pf_counter_u64_fetch(&tail->bytes[0]));
1136 				pf_counter_u64_add_protected(&rule->bytes[1],
1137 				    pf_counter_u64_fetch(&tail->bytes[1]));
1138 				pf_counter_u64_critical_exit();
1139 			}
1140 		}
1141 	}
1142 
1143 	rs->rules[rs_num].inactive.ptr = old_rules;
1144 	rs->rules[rs_num].inactive.ptr_array = old_array;
1145 	rs->rules[rs_num].inactive.rcount = old_rcount;
1146 
1147 	rs->rules[rs_num].active.ticket =
1148 	    rs->rules[rs_num].inactive.ticket;
1149 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1150 
1151 	/* Purge the old rule list. */
1152 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1153 		pf_unlink_rule(old_rules, rule);
1154 	if (rs->rules[rs_num].inactive.ptr_array)
1155 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1156 	rs->rules[rs_num].inactive.ptr_array = NULL;
1157 	rs->rules[rs_num].inactive.rcount = 0;
1158 	rs->rules[rs_num].inactive.open = 0;
1159 	pf_remove_if_empty_kruleset(rs);
1160 
1161 	return (0);
1162 }
1163 
1164 static int
1165 pf_setup_pfsync_matching(struct pf_kruleset *rs)
1166 {
1167 	MD5_CTX			 ctx;
1168 	struct pf_krule		*rule;
1169 	int			 rs_cnt;
1170 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1171 
1172 	MD5Init(&ctx);
1173 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1174 		/* XXX PF_RULESET_SCRUB as well? */
1175 		if (rs_cnt == PF_RULESET_SCRUB)
1176 			continue;
1177 
1178 		if (rs->rules[rs_cnt].inactive.ptr_array)
1179 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1180 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1181 
1182 		if (rs->rules[rs_cnt].inactive.rcount) {
1183 			rs->rules[rs_cnt].inactive.ptr_array =
1184 			    malloc(sizeof(caddr_t) *
1185 			    rs->rules[rs_cnt].inactive.rcount,
1186 			    M_TEMP, M_NOWAIT);
1187 
1188 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1189 				return (ENOMEM);
1190 		}
1191 
1192 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1193 		    entries) {
1194 			pf_hash_rule(&ctx, rule);
1195 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1196 		}
1197 	}
1198 
1199 	MD5Final(digest, &ctx);
1200 	memcpy(V_pf_status.pf_chksum, digest, sizeof(V_pf_status.pf_chksum));
1201 	return (0);
1202 }
1203 
1204 static int
1205 pf_addr_setup(struct pf_kruleset *ruleset, struct pf_addr_wrap *addr,
1206     sa_family_t af)
1207 {
1208 	int error = 0;
1209 
1210 	switch (addr->type) {
1211 	case PF_ADDR_TABLE:
1212 		addr->p.tbl = pfr_attach_table(ruleset, addr->v.tblname);
1213 		if (addr->p.tbl == NULL)
1214 			error = ENOMEM;
1215 		break;
1216 	case PF_ADDR_DYNIFTL:
1217 		error = pfi_dynaddr_setup(addr, af);
1218 		break;
1219 	}
1220 
1221 	return (error);
1222 }
1223 
1224 static void
1225 pf_addr_copyout(struct pf_addr_wrap *addr)
1226 {
1227 
1228 	switch (addr->type) {
1229 	case PF_ADDR_DYNIFTL:
1230 		pfi_dynaddr_copyout(addr);
1231 		break;
1232 	case PF_ADDR_TABLE:
1233 		pf_tbladdr_copyout(addr);
1234 		break;
1235 	}
1236 }
1237 
1238 static void
1239 pf_src_node_copy(const struct pf_ksrc_node *in, struct pf_src_node *out)
1240 {
1241 	int	secs = time_uptime, diff;
1242 
1243 	bzero(out, sizeof(struct pf_src_node));
1244 
1245 	bcopy(&in->addr, &out->addr, sizeof(struct pf_addr));
1246 	bcopy(&in->raddr, &out->raddr, sizeof(struct pf_addr));
1247 
1248 	if (in->rule.ptr != NULL)
1249 		out->rule.nr = in->rule.ptr->nr;
1250 
1251 	for (int i = 0; i < 2; i++) {
1252 		out->bytes[i] = counter_u64_fetch(in->bytes[i]);
1253 		out->packets[i] = counter_u64_fetch(in->packets[i]);
1254 	}
1255 
1256 	out->states = in->states;
1257 	out->conn = in->conn;
1258 	out->af = in->af;
1259 	out->ruletype = in->ruletype;
1260 
1261 	out->creation = secs - in->creation;
1262 	if (out->expire > secs)
1263 		out->expire -= secs;
1264 	else
1265 		out->expire = 0;
1266 
1267 	/* Adjust the connection rate estimate. */
1268 	diff = secs - in->conn_rate.last;
1269 	if (diff >= in->conn_rate.seconds)
1270 		out->conn_rate.count = 0;
1271 	else
1272 		out->conn_rate.count -=
1273 		    in->conn_rate.count * diff /
1274 		    in->conn_rate.seconds;
1275 }
1276 
1277 #ifdef ALTQ
1278 /*
1279  * Handle export of struct pf_kaltq to user binaries that may be using any
1280  * version of struct pf_altq.
1281  */
1282 static int
1283 pf_export_kaltq(struct pf_altq *q, struct pfioc_altq_v1 *pa, size_t ioc_size)
1284 {
1285 	u_int32_t version;
1286 
1287 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1288 		version = 0;
1289 	else
1290 		version = pa->version;
1291 
1292 	if (version > PFIOC_ALTQ_VERSION)
1293 		return (EINVAL);
1294 
1295 #define ASSIGN(x) exported_q->x = q->x
1296 #define COPY(x) \
1297 	bcopy(&q->x, &exported_q->x, min(sizeof(q->x), sizeof(exported_q->x)))
1298 #define SATU16(x) (u_int32_t)uqmin((x), USHRT_MAX)
1299 #define SATU32(x) (u_int32_t)uqmin((x), UINT_MAX)
1300 
1301 	switch (version) {
1302 	case 0: {
1303 		struct pf_altq_v0 *exported_q =
1304 		    &((struct pfioc_altq_v0 *)pa)->altq;
1305 
1306 		COPY(ifname);
1307 
1308 		ASSIGN(scheduler);
1309 		ASSIGN(tbrsize);
1310 		exported_q->tbrsize = SATU16(q->tbrsize);
1311 		exported_q->ifbandwidth = SATU32(q->ifbandwidth);
1312 
1313 		COPY(qname);
1314 		COPY(parent);
1315 		ASSIGN(parent_qid);
1316 		exported_q->bandwidth = SATU32(q->bandwidth);
1317 		ASSIGN(priority);
1318 		ASSIGN(local_flags);
1319 
1320 		ASSIGN(qlimit);
1321 		ASSIGN(flags);
1322 
1323 		if (q->scheduler == ALTQT_HFSC) {
1324 #define ASSIGN_OPT(x) exported_q->pq_u.hfsc_opts.x = q->pq_u.hfsc_opts.x
1325 #define ASSIGN_OPT_SATU32(x) exported_q->pq_u.hfsc_opts.x = \
1326 			    SATU32(q->pq_u.hfsc_opts.x)
1327 
1328 			ASSIGN_OPT_SATU32(rtsc_m1);
1329 			ASSIGN_OPT(rtsc_d);
1330 			ASSIGN_OPT_SATU32(rtsc_m2);
1331 
1332 			ASSIGN_OPT_SATU32(lssc_m1);
1333 			ASSIGN_OPT(lssc_d);
1334 			ASSIGN_OPT_SATU32(lssc_m2);
1335 
1336 			ASSIGN_OPT_SATU32(ulsc_m1);
1337 			ASSIGN_OPT(ulsc_d);
1338 			ASSIGN_OPT_SATU32(ulsc_m2);
1339 
1340 			ASSIGN_OPT(flags);
1341 
1342 #undef ASSIGN_OPT
1343 #undef ASSIGN_OPT_SATU32
1344 		} else
1345 			COPY(pq_u);
1346 
1347 		ASSIGN(qid);
1348 		break;
1349 	}
1350 	case 1:	{
1351 		struct pf_altq_v1 *exported_q =
1352 		    &((struct pfioc_altq_v1 *)pa)->altq;
1353 
1354 		COPY(ifname);
1355 
1356 		ASSIGN(scheduler);
1357 		ASSIGN(tbrsize);
1358 		ASSIGN(ifbandwidth);
1359 
1360 		COPY(qname);
1361 		COPY(parent);
1362 		ASSIGN(parent_qid);
1363 		ASSIGN(bandwidth);
1364 		ASSIGN(priority);
1365 		ASSIGN(local_flags);
1366 
1367 		ASSIGN(qlimit);
1368 		ASSIGN(flags);
1369 		COPY(pq_u);
1370 
1371 		ASSIGN(qid);
1372 		break;
1373 	}
1374 	default:
1375 		panic("%s: unhandled struct pfioc_altq version", __func__);
1376 		break;
1377 	}
1378 
1379 #undef ASSIGN
1380 #undef COPY
1381 #undef SATU16
1382 #undef SATU32
1383 
1384 	return (0);
1385 }
1386 
1387 /*
1388  * Handle import to struct pf_kaltq of struct pf_altq from user binaries
1389  * that may be using any version of it.
1390  */
1391 static int
1392 pf_import_kaltq(struct pfioc_altq_v1 *pa, struct pf_altq *q, size_t ioc_size)
1393 {
1394 	u_int32_t version;
1395 
1396 	if (ioc_size == sizeof(struct pfioc_altq_v0))
1397 		version = 0;
1398 	else
1399 		version = pa->version;
1400 
1401 	if (version > PFIOC_ALTQ_VERSION)
1402 		return (EINVAL);
1403 
1404 #define ASSIGN(x) q->x = imported_q->x
1405 #define COPY(x) \
1406 	bcopy(&imported_q->x, &q->x, min(sizeof(imported_q->x), sizeof(q->x)))
1407 
1408 	switch (version) {
1409 	case 0: {
1410 		struct pf_altq_v0 *imported_q =
1411 		    &((struct pfioc_altq_v0 *)pa)->altq;
1412 
1413 		COPY(ifname);
1414 
1415 		ASSIGN(scheduler);
1416 		ASSIGN(tbrsize); /* 16-bit -> 32-bit */
1417 		ASSIGN(ifbandwidth); /* 32-bit -> 64-bit */
1418 
1419 		COPY(qname);
1420 		COPY(parent);
1421 		ASSIGN(parent_qid);
1422 		ASSIGN(bandwidth); /* 32-bit -> 64-bit */
1423 		ASSIGN(priority);
1424 		ASSIGN(local_flags);
1425 
1426 		ASSIGN(qlimit);
1427 		ASSIGN(flags);
1428 
1429 		if (imported_q->scheduler == ALTQT_HFSC) {
1430 #define ASSIGN_OPT(x) q->pq_u.hfsc_opts.x = imported_q->pq_u.hfsc_opts.x
1431 
1432 			/*
1433 			 * The m1 and m2 parameters are being copied from
1434 			 * 32-bit to 64-bit.
1435 			 */
1436 			ASSIGN_OPT(rtsc_m1);
1437 			ASSIGN_OPT(rtsc_d);
1438 			ASSIGN_OPT(rtsc_m2);
1439 
1440 			ASSIGN_OPT(lssc_m1);
1441 			ASSIGN_OPT(lssc_d);
1442 			ASSIGN_OPT(lssc_m2);
1443 
1444 			ASSIGN_OPT(ulsc_m1);
1445 			ASSIGN_OPT(ulsc_d);
1446 			ASSIGN_OPT(ulsc_m2);
1447 
1448 			ASSIGN_OPT(flags);
1449 
1450 #undef ASSIGN_OPT
1451 		} else
1452 			COPY(pq_u);
1453 
1454 		ASSIGN(qid);
1455 		break;
1456 	}
1457 	case 1: {
1458 		struct pf_altq_v1 *imported_q =
1459 		    &((struct pfioc_altq_v1 *)pa)->altq;
1460 
1461 		COPY(ifname);
1462 
1463 		ASSIGN(scheduler);
1464 		ASSIGN(tbrsize);
1465 		ASSIGN(ifbandwidth);
1466 
1467 		COPY(qname);
1468 		COPY(parent);
1469 		ASSIGN(parent_qid);
1470 		ASSIGN(bandwidth);
1471 		ASSIGN(priority);
1472 		ASSIGN(local_flags);
1473 
1474 		ASSIGN(qlimit);
1475 		ASSIGN(flags);
1476 		COPY(pq_u);
1477 
1478 		ASSIGN(qid);
1479 		break;
1480 	}
1481 	default:
1482 		panic("%s: unhandled struct pfioc_altq version", __func__);
1483 		break;
1484 	}
1485 
1486 #undef ASSIGN
1487 #undef COPY
1488 
1489 	return (0);
1490 }
1491 
1492 static struct pf_altq *
1493 pf_altq_get_nth_active(u_int32_t n)
1494 {
1495 	struct pf_altq		*altq;
1496 	u_int32_t		 nr;
1497 
1498 	nr = 0;
1499 	TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
1500 		if (nr == n)
1501 			return (altq);
1502 		nr++;
1503 	}
1504 
1505 	TAILQ_FOREACH(altq, V_pf_altqs_active, entries) {
1506 		if (nr == n)
1507 			return (altq);
1508 		nr++;
1509 	}
1510 
1511 	return (NULL);
1512 }
1513 #endif /* ALTQ */
1514 
1515 void
1516 pf_krule_free(struct pf_krule *rule)
1517 {
1518 #ifdef PF_WANT_32_TO_64_COUNTER
1519 	bool wowned;
1520 #endif
1521 
1522 	if (rule == NULL)
1523 		return;
1524 
1525 #ifdef PF_WANT_32_TO_64_COUNTER
1526 	if (rule->allrulelinked) {
1527 		wowned = PF_RULES_WOWNED();
1528 		if (!wowned)
1529 			PF_RULES_WLOCK();
1530 		LIST_REMOVE(rule, allrulelist);
1531 		V_pf_allrulecount--;
1532 		if (!wowned)
1533 			PF_RULES_WUNLOCK();
1534 	}
1535 #endif
1536 
1537 	pf_counter_u64_deinit(&rule->evaluations);
1538 	for (int i = 0; i < 2; i++) {
1539 		pf_counter_u64_deinit(&rule->packets[i]);
1540 		pf_counter_u64_deinit(&rule->bytes[i]);
1541 	}
1542 	counter_u64_free(rule->states_cur);
1543 	counter_u64_free(rule->states_tot);
1544 	counter_u64_free(rule->src_nodes);
1545 	free(rule, M_PFRULE);
1546 }
1547 
1548 static void
1549 pf_kpooladdr_to_pooladdr(const struct pf_kpooladdr *kpool,
1550     struct pf_pooladdr *pool)
1551 {
1552 
1553 	bzero(pool, sizeof(*pool));
1554 	bcopy(&kpool->addr, &pool->addr, sizeof(pool->addr));
1555 	strlcpy(pool->ifname, kpool->ifname, sizeof(pool->ifname));
1556 }
1557 
1558 static int
1559 pf_pooladdr_to_kpooladdr(const struct pf_pooladdr *pool,
1560     struct pf_kpooladdr *kpool)
1561 {
1562 	int ret;
1563 
1564 	bzero(kpool, sizeof(*kpool));
1565 	bcopy(&pool->addr, &kpool->addr, sizeof(kpool->addr));
1566 	ret = pf_user_strcpy(kpool->ifname, pool->ifname,
1567 	    sizeof(kpool->ifname));
1568 	return (ret);
1569 }
1570 
1571 static void
1572 pf_kpool_to_pool(const struct pf_kpool *kpool, struct pf_pool *pool)
1573 {
1574 	bzero(pool, sizeof(*pool));
1575 
1576 	bcopy(&kpool->key, &pool->key, sizeof(pool->key));
1577 	bcopy(&kpool->counter, &pool->counter, sizeof(pool->counter));
1578 
1579 	pool->tblidx = kpool->tblidx;
1580 	pool->proxy_port[0] = kpool->proxy_port[0];
1581 	pool->proxy_port[1] = kpool->proxy_port[1];
1582 	pool->opts = kpool->opts;
1583 }
1584 
1585 static int
1586 pf_pool_to_kpool(const struct pf_pool *pool, struct pf_kpool *kpool)
1587 {
1588 	_Static_assert(sizeof(pool->key) == sizeof(kpool->key), "");
1589 	_Static_assert(sizeof(pool->counter) == sizeof(kpool->counter), "");
1590 
1591 	bzero(kpool, sizeof(*kpool));
1592 
1593 	bcopy(&pool->key, &kpool->key, sizeof(kpool->key));
1594 	bcopy(&pool->counter, &kpool->counter, sizeof(kpool->counter));
1595 
1596 	kpool->tblidx = pool->tblidx;
1597 	kpool->proxy_port[0] = pool->proxy_port[0];
1598 	kpool->proxy_port[1] = pool->proxy_port[1];
1599 	kpool->opts = pool->opts;
1600 
1601 	return (0);
1602 }
1603 
1604 static void
1605 pf_krule_to_rule(struct pf_krule *krule, struct pf_rule *rule)
1606 {
1607 
1608 	bzero(rule, sizeof(*rule));
1609 
1610 	bcopy(&krule->src, &rule->src, sizeof(rule->src));
1611 	bcopy(&krule->dst, &rule->dst, sizeof(rule->dst));
1612 
1613 	for (int i = 0; i < PF_SKIP_COUNT; ++i) {
1614 		if (rule->skip[i].ptr == NULL)
1615 			rule->skip[i].nr = -1;
1616 		else
1617 			rule->skip[i].nr = krule->skip[i].ptr->nr;
1618 	}
1619 
1620 	strlcpy(rule->label, krule->label[0], sizeof(rule->label));
1621 	strlcpy(rule->ifname, krule->ifname, sizeof(rule->ifname));
1622 	strlcpy(rule->qname, krule->qname, sizeof(rule->qname));
1623 	strlcpy(rule->pqname, krule->pqname, sizeof(rule->pqname));
1624 	strlcpy(rule->tagname, krule->tagname, sizeof(rule->tagname));
1625 	strlcpy(rule->match_tagname, krule->match_tagname,
1626 	    sizeof(rule->match_tagname));
1627 	strlcpy(rule->overload_tblname, krule->overload_tblname,
1628 	    sizeof(rule->overload_tblname));
1629 
1630 	pf_kpool_to_pool(&krule->rpool, &rule->rpool);
1631 
1632 	rule->evaluations = pf_counter_u64_fetch(&krule->evaluations);
1633 	for (int i = 0; i < 2; i++) {
1634 		rule->packets[i] = pf_counter_u64_fetch(&krule->packets[i]);
1635 		rule->bytes[i] = pf_counter_u64_fetch(&krule->bytes[i]);
1636 	}
1637 
1638 	/* kif, anchor, overload_tbl are not copied over. */
1639 
1640 	rule->os_fingerprint = krule->os_fingerprint;
1641 
1642 	rule->rtableid = krule->rtableid;
1643 	bcopy(krule->timeout, rule->timeout, sizeof(krule->timeout));
1644 	rule->max_states = krule->max_states;
1645 	rule->max_src_nodes = krule->max_src_nodes;
1646 	rule->max_src_states = krule->max_src_states;
1647 	rule->max_src_conn = krule->max_src_conn;
1648 	rule->max_src_conn_rate.limit = krule->max_src_conn_rate.limit;
1649 	rule->max_src_conn_rate.seconds = krule->max_src_conn_rate.seconds;
1650 	rule->qid = krule->qid;
1651 	rule->pqid = krule->pqid;
1652 	rule->nr = krule->nr;
1653 	rule->prob = krule->prob;
1654 	rule->cuid = krule->cuid;
1655 	rule->cpid = krule->cpid;
1656 
1657 	rule->return_icmp = krule->return_icmp;
1658 	rule->return_icmp6 = krule->return_icmp6;
1659 	rule->max_mss = krule->max_mss;
1660 	rule->tag = krule->tag;
1661 	rule->match_tag = krule->match_tag;
1662 	rule->scrub_flags = krule->scrub_flags;
1663 
1664 	bcopy(&krule->uid, &rule->uid, sizeof(krule->uid));
1665 	bcopy(&krule->gid, &rule->gid, sizeof(krule->gid));
1666 
1667 	rule->rule_flag = krule->rule_flag;
1668 	rule->action = krule->action;
1669 	rule->direction = krule->direction;
1670 	rule->log = krule->log;
1671 	rule->logif = krule->logif;
1672 	rule->quick = krule->quick;
1673 	rule->ifnot = krule->ifnot;
1674 	rule->match_tag_not = krule->match_tag_not;
1675 	rule->natpass = krule->natpass;
1676 
1677 	rule->keep_state = krule->keep_state;
1678 	rule->af = krule->af;
1679 	rule->proto = krule->proto;
1680 	rule->type = krule->type;
1681 	rule->code = krule->code;
1682 	rule->flags = krule->flags;
1683 	rule->flagset = krule->flagset;
1684 	rule->min_ttl = krule->min_ttl;
1685 	rule->allow_opts = krule->allow_opts;
1686 	rule->rt = krule->rt;
1687 	rule->return_ttl = krule->return_ttl;
1688 	rule->tos = krule->tos;
1689 	rule->set_tos = krule->set_tos;
1690 	rule->anchor_relative = krule->anchor_relative;
1691 	rule->anchor_wildcard = krule->anchor_wildcard;
1692 
1693 	rule->flush = krule->flush;
1694 	rule->prio = krule->prio;
1695 	rule->set_prio[0] = krule->set_prio[0];
1696 	rule->set_prio[1] = krule->set_prio[1];
1697 
1698 	bcopy(&krule->divert, &rule->divert, sizeof(krule->divert));
1699 
1700 	rule->u_states_cur = counter_u64_fetch(krule->states_cur);
1701 	rule->u_states_tot = counter_u64_fetch(krule->states_tot);
1702 	rule->u_src_nodes = counter_u64_fetch(krule->src_nodes);
1703 }
1704 
1705 static int
1706 pf_rule_to_krule(const struct pf_rule *rule, struct pf_krule *krule)
1707 {
1708 	int ret;
1709 
1710 #ifndef INET
1711 	if (rule->af == AF_INET) {
1712 		return (EAFNOSUPPORT);
1713 	}
1714 #endif /* INET */
1715 #ifndef INET6
1716 	if (rule->af == AF_INET6) {
1717 		return (EAFNOSUPPORT);
1718 	}
1719 #endif /* INET6 */
1720 
1721 	ret = pf_check_rule_addr(&rule->src);
1722 	if (ret != 0)
1723 		return (ret);
1724 	ret = pf_check_rule_addr(&rule->dst);
1725 	if (ret != 0)
1726 		return (ret);
1727 
1728 	bzero(krule, sizeof(*krule));
1729 
1730 	bcopy(&rule->src, &krule->src, sizeof(rule->src));
1731 	bcopy(&rule->dst, &krule->dst, sizeof(rule->dst));
1732 
1733 	ret = pf_user_strcpy(krule->label[0], rule->label, sizeof(rule->label));
1734 	if (ret != 0)
1735 		return (ret);
1736 	ret = pf_user_strcpy(krule->ifname, rule->ifname, sizeof(rule->ifname));
1737 	if (ret != 0)
1738 		return (ret);
1739 	ret = pf_user_strcpy(krule->qname, rule->qname, sizeof(rule->qname));
1740 	if (ret != 0)
1741 		return (ret);
1742 	ret = pf_user_strcpy(krule->pqname, rule->pqname, sizeof(rule->pqname));
1743 	if (ret != 0)
1744 		return (ret);
1745 	ret = pf_user_strcpy(krule->tagname, rule->tagname,
1746 	    sizeof(rule->tagname));
1747 	if (ret != 0)
1748 		return (ret);
1749 	ret = pf_user_strcpy(krule->match_tagname, rule->match_tagname,
1750 	    sizeof(rule->match_tagname));
1751 	if (ret != 0)
1752 		return (ret);
1753 	ret = pf_user_strcpy(krule->overload_tblname, rule->overload_tblname,
1754 	    sizeof(rule->overload_tblname));
1755 	if (ret != 0)
1756 		return (ret);
1757 
1758 	ret = pf_pool_to_kpool(&rule->rpool, &krule->rpool);
1759 	if (ret != 0)
1760 		return (ret);
1761 
1762 	/* Don't allow userspace to set evaulations, packets or bytes. */
1763 	/* kif, anchor, overload_tbl are not copied over. */
1764 
1765 	krule->os_fingerprint = rule->os_fingerprint;
1766 
1767 	krule->rtableid = rule->rtableid;
1768 	bcopy(rule->timeout, krule->timeout, sizeof(krule->timeout));
1769 	krule->max_states = rule->max_states;
1770 	krule->max_src_nodes = rule->max_src_nodes;
1771 	krule->max_src_states = rule->max_src_states;
1772 	krule->max_src_conn = rule->max_src_conn;
1773 	krule->max_src_conn_rate.limit = rule->max_src_conn_rate.limit;
1774 	krule->max_src_conn_rate.seconds = rule->max_src_conn_rate.seconds;
1775 	krule->qid = rule->qid;
1776 	krule->pqid = rule->pqid;
1777 	krule->nr = rule->nr;
1778 	krule->prob = rule->prob;
1779 	krule->cuid = rule->cuid;
1780 	krule->cpid = rule->cpid;
1781 
1782 	krule->return_icmp = rule->return_icmp;
1783 	krule->return_icmp6 = rule->return_icmp6;
1784 	krule->max_mss = rule->max_mss;
1785 	krule->tag = rule->tag;
1786 	krule->match_tag = rule->match_tag;
1787 	krule->scrub_flags = rule->scrub_flags;
1788 
1789 	bcopy(&rule->uid, &krule->uid, sizeof(krule->uid));
1790 	bcopy(&rule->gid, &krule->gid, sizeof(krule->gid));
1791 
1792 	krule->rule_flag = rule->rule_flag;
1793 	krule->action = rule->action;
1794 	krule->direction = rule->direction;
1795 	krule->log = rule->log;
1796 	krule->logif = rule->logif;
1797 	krule->quick = rule->quick;
1798 	krule->ifnot = rule->ifnot;
1799 	krule->match_tag_not = rule->match_tag_not;
1800 	krule->natpass = rule->natpass;
1801 
1802 	krule->keep_state = rule->keep_state;
1803 	krule->af = rule->af;
1804 	krule->proto = rule->proto;
1805 	krule->type = rule->type;
1806 	krule->code = rule->code;
1807 	krule->flags = rule->flags;
1808 	krule->flagset = rule->flagset;
1809 	krule->min_ttl = rule->min_ttl;
1810 	krule->allow_opts = rule->allow_opts;
1811 	krule->rt = rule->rt;
1812 	krule->return_ttl = rule->return_ttl;
1813 	krule->tos = rule->tos;
1814 	krule->set_tos = rule->set_tos;
1815 
1816 	krule->flush = rule->flush;
1817 	krule->prio = rule->prio;
1818 	krule->set_prio[0] = rule->set_prio[0];
1819 	krule->set_prio[1] = rule->set_prio[1];
1820 
1821 	bcopy(&rule->divert, &krule->divert, sizeof(krule->divert));
1822 
1823 	return (0);
1824 }
1825 
1826 static int
1827 pf_state_kill_to_kstate_kill(const struct pfioc_state_kill *psk,
1828     struct pf_kstate_kill *kill)
1829 {
1830 	int ret;
1831 
1832 	bzero(kill, sizeof(*kill));
1833 
1834 	bcopy(&psk->psk_pfcmp, &kill->psk_pfcmp, sizeof(kill->psk_pfcmp));
1835 	kill->psk_af = psk->psk_af;
1836 	kill->psk_proto = psk->psk_proto;
1837 	bcopy(&psk->psk_src, &kill->psk_src, sizeof(kill->psk_src));
1838 	bcopy(&psk->psk_dst, &kill->psk_dst, sizeof(kill->psk_dst));
1839 	ret = pf_user_strcpy(kill->psk_ifname, psk->psk_ifname,
1840 	    sizeof(kill->psk_ifname));
1841 	if (ret != 0)
1842 		return (ret);
1843 	ret = pf_user_strcpy(kill->psk_label, psk->psk_label,
1844 	    sizeof(kill->psk_label));
1845 	if (ret != 0)
1846 		return (ret);
1847 
1848 	return (0);
1849 }
1850 
1851 static int
1852 pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket,
1853     uint32_t pool_ticket, const char *anchor, const char *anchor_call,
1854     struct thread *td)
1855 {
1856 	struct pf_kruleset	*ruleset;
1857 	struct pf_krule		*tail;
1858 	struct pf_kpooladdr	*pa;
1859 	struct pfi_kkif		*kif = NULL;
1860 	int			 rs_num;
1861 	int			 error = 0;
1862 
1863 	if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) {
1864 		error = EINVAL;
1865 		goto errout_unlocked;
1866 	}
1867 
1868 #define	ERROUT(x)	ERROUT_FUNCTION(errout, x)
1869 
1870 	if (rule->ifname[0])
1871 		kif = pf_kkif_create(M_WAITOK);
1872 	pf_counter_u64_init(&rule->evaluations, M_WAITOK);
1873 	for (int i = 0; i < 2; i++) {
1874 		pf_counter_u64_init(&rule->packets[i], M_WAITOK);
1875 		pf_counter_u64_init(&rule->bytes[i], M_WAITOK);
1876 	}
1877 	rule->states_cur = counter_u64_alloc(M_WAITOK);
1878 	rule->states_tot = counter_u64_alloc(M_WAITOK);
1879 	rule->src_nodes = counter_u64_alloc(M_WAITOK);
1880 	rule->cuid = td->td_ucred->cr_ruid;
1881 	rule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
1882 	TAILQ_INIT(&rule->rpool.list);
1883 
1884 	PF_RULES_WLOCK();
1885 #ifdef PF_WANT_32_TO_64_COUNTER
1886 	LIST_INSERT_HEAD(&V_pf_allrulelist, rule, allrulelist);
1887 	MPASS(!rule->allrulelinked);
1888 	rule->allrulelinked = true;
1889 	V_pf_allrulecount++;
1890 #endif
1891 	ruleset = pf_find_kruleset(anchor);
1892 	if (ruleset == NULL)
1893 		ERROUT(EINVAL);
1894 	rs_num = pf_get_ruleset_number(rule->action);
1895 	if (rs_num >= PF_RULESET_MAX)
1896 		ERROUT(EINVAL);
1897 	if (ticket != ruleset->rules[rs_num].inactive.ticket) {
1898 		DPFPRINTF(PF_DEBUG_MISC,
1899 		    ("ticket: %d != [%d]%d\n", ticket, rs_num,
1900 		    ruleset->rules[rs_num].inactive.ticket));
1901 		ERROUT(EBUSY);
1902 	}
1903 	if (pool_ticket != V_ticket_pabuf) {
1904 		DPFPRINTF(PF_DEBUG_MISC,
1905 		    ("pool_ticket: %d != %d\n", pool_ticket,
1906 		    V_ticket_pabuf));
1907 		ERROUT(EBUSY);
1908 	}
1909 
1910 	tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1911 	    pf_krulequeue);
1912 	if (tail)
1913 		rule->nr = tail->nr + 1;
1914 	else
1915 		rule->nr = 0;
1916 	if (rule->ifname[0]) {
1917 		rule->kif = pfi_kkif_attach(kif, rule->ifname);
1918 		kif = NULL;
1919 		pfi_kkif_ref(rule->kif);
1920 	} else
1921 		rule->kif = NULL;
1922 
1923 	if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs)
1924 		error = EBUSY;
1925 
1926 #ifdef ALTQ
1927 	/* set queue IDs */
1928 	if (rule->qname[0] != 0) {
1929 		if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1930 			error = EBUSY;
1931 		else if (rule->pqname[0] != 0) {
1932 			if ((rule->pqid =
1933 			    pf_qname2qid(rule->pqname)) == 0)
1934 				error = EBUSY;
1935 		} else
1936 			rule->pqid = rule->qid;
1937 	}
1938 #endif
1939 	if (rule->tagname[0])
1940 		if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1941 			error = EBUSY;
1942 	if (rule->match_tagname[0])
1943 		if ((rule->match_tag =
1944 		    pf_tagname2tag(rule->match_tagname)) == 0)
1945 			error = EBUSY;
1946 	if (rule->rt && !rule->direction)
1947 		error = EINVAL;
1948 	if (!rule->log)
1949 		rule->logif = 0;
1950 	if (rule->logif >= PFLOGIFS_MAX)
1951 		error = EINVAL;
1952 	if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1953 		error = ENOMEM;
1954 	if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1955 		error = ENOMEM;
1956 	if (pf_kanchor_setup(rule, ruleset, anchor_call))
1957 		error = EINVAL;
1958 	if (rule->scrub_flags & PFSTATE_SETPRIO &&
1959 	    (rule->set_prio[0] > PF_PRIO_MAX ||
1960 	    rule->set_prio[1] > PF_PRIO_MAX))
1961 		error = EINVAL;
1962 	TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
1963 		if (pa->addr.type == PF_ADDR_TABLE) {
1964 			pa->addr.p.tbl = pfr_attach_table(ruleset,
1965 			    pa->addr.v.tblname);
1966 			if (pa->addr.p.tbl == NULL)
1967 				error = ENOMEM;
1968 		}
1969 
1970 	rule->overload_tbl = NULL;
1971 	if (rule->overload_tblname[0]) {
1972 		if ((rule->overload_tbl = pfr_attach_table(ruleset,
1973 		    rule->overload_tblname)) == NULL)
1974 			error = EINVAL;
1975 		else
1976 			rule->overload_tbl->pfrkt_flags |=
1977 			    PFR_TFLAG_ACTIVE;
1978 	}
1979 
1980 	pf_mv_kpool(&V_pf_pabuf, &rule->rpool.list);
1981 	if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1982 	    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1983 	    (rule->rt > PF_NOPFROUTE)) &&
1984 	    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1985 		error = EINVAL;
1986 
1987 	if (error) {
1988 		pf_free_rule(rule);
1989 		rule = NULL;
1990 		ERROUT(error);
1991 	}
1992 
1993 	rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1994 	pf_counter_u64_zero(&rule->evaluations);
1995 	for (int i = 0; i < 2; i++) {
1996 		pf_counter_u64_zero(&rule->packets[i]);
1997 		pf_counter_u64_zero(&rule->bytes[i]);
1998 	}
1999 	TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
2000 	    rule, entries);
2001 	ruleset->rules[rs_num].inactive.rcount++;
2002 	PF_RULES_WUNLOCK();
2003 
2004 	return (0);
2005 
2006 #undef ERROUT
2007 errout:
2008 	PF_RULES_WUNLOCK();
2009 errout_unlocked:
2010 	pf_kkif_free(kif);
2011 	pf_krule_free(rule);
2012 	return (error);
2013 }
2014 
2015 static bool
2016 pf_label_match(const struct pf_krule *rule, const char *label)
2017 {
2018 	int i = 0;
2019 
2020 	while (*rule->label[i]) {
2021 		if (strcmp(rule->label[i], label) == 0)
2022 			return (true);
2023 		i++;
2024 	}
2025 
2026 	return (false);
2027 }
2028 
2029 static unsigned int
2030 pf_kill_matching_state(struct pf_state_key_cmp *key, int dir)
2031 {
2032 	struct pf_kstate *match;
2033 	int more = 0;
2034 	unsigned int killed = 0;
2035 
2036 	/* Call with unlocked hashrow */
2037 
2038 	match = pf_find_state_all(key, dir, &more);
2039 	if (match && !more) {
2040 		pf_unlink_state(match, 0);
2041 		killed++;
2042 	}
2043 
2044 	return (killed);
2045 }
2046 
2047 static int
2048 pf_killstates_row(struct pf_kstate_kill *psk, struct pf_idhash *ih)
2049 {
2050 	struct pf_kstate	*s;
2051 	struct pf_state_key	*sk;
2052 	struct pf_addr		*srcaddr, *dstaddr;
2053 	struct pf_state_key_cmp	 match_key;
2054 	int			 idx, killed = 0;
2055 	unsigned int		 dir;
2056 	u_int16_t		 srcport, dstport;
2057 	struct pfi_kkif		*kif;
2058 
2059 relock_DIOCKILLSTATES:
2060 	PF_HASHROW_LOCK(ih);
2061 	LIST_FOREACH(s, &ih->states, entry) {
2062 		/* For floating states look at the original kif. */
2063 		kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
2064 
2065 		sk = s->key[PF_SK_WIRE];
2066 		if (s->direction == PF_OUT) {
2067 			srcaddr = &sk->addr[1];
2068 			dstaddr = &sk->addr[0];
2069 			srcport = sk->port[1];
2070 			dstport = sk->port[0];
2071 		} else {
2072 			srcaddr = &sk->addr[0];
2073 			dstaddr = &sk->addr[1];
2074 			srcport = sk->port[0];
2075 			dstport = sk->port[1];
2076 		}
2077 
2078 		if (psk->psk_af && sk->af != psk->psk_af)
2079 			continue;
2080 
2081 		if (psk->psk_proto && psk->psk_proto != sk->proto)
2082 			continue;
2083 
2084 		if (! PF_MATCHA(psk->psk_src.neg, &psk->psk_src.addr.v.a.addr,
2085 		    &psk->psk_src.addr.v.a.mask, srcaddr, sk->af))
2086 			continue;
2087 
2088 		if (! PF_MATCHA(psk->psk_dst.neg, &psk->psk_dst.addr.v.a.addr,
2089 		    &psk->psk_dst.addr.v.a.mask, dstaddr, sk->af))
2090 			continue;
2091 
2092 		if (!  PF_MATCHA(psk->psk_rt_addr.neg,
2093 		    &psk->psk_rt_addr.addr.v.a.addr,
2094 		    &psk->psk_rt_addr.addr.v.a.mask,
2095 		    &s->rt_addr, sk->af))
2096 			continue;
2097 
2098 		if (psk->psk_src.port_op != 0 &&
2099 		    ! pf_match_port(psk->psk_src.port_op,
2100 		    psk->psk_src.port[0], psk->psk_src.port[1], srcport))
2101 			continue;
2102 
2103 		if (psk->psk_dst.port_op != 0 &&
2104 		    ! pf_match_port(psk->psk_dst.port_op,
2105 		    psk->psk_dst.port[0], psk->psk_dst.port[1], dstport))
2106 			continue;
2107 
2108 		if (psk->psk_label[0] &&
2109 		    ! pf_label_match(s->rule.ptr, psk->psk_label))
2110 			continue;
2111 
2112 		if (psk->psk_ifname[0] && strcmp(psk->psk_ifname,
2113 		    kif->pfik_name))
2114 			continue;
2115 
2116 		if (psk->psk_kill_match) {
2117 			/* Create the key to find matching states, with lock
2118 			 * held. */
2119 
2120 			bzero(&match_key, sizeof(match_key));
2121 
2122 			if (s->direction == PF_OUT) {
2123 				dir = PF_IN;
2124 				idx = PF_SK_STACK;
2125 			} else {
2126 				dir = PF_OUT;
2127 				idx = PF_SK_WIRE;
2128 			}
2129 
2130 			match_key.af = s->key[idx]->af;
2131 			match_key.proto = s->key[idx]->proto;
2132 			PF_ACPY(&match_key.addr[0],
2133 			    &s->key[idx]->addr[1], match_key.af);
2134 			match_key.port[0] = s->key[idx]->port[1];
2135 			PF_ACPY(&match_key.addr[1],
2136 			    &s->key[idx]->addr[0], match_key.af);
2137 			match_key.port[1] = s->key[idx]->port[0];
2138 		}
2139 
2140 		pf_unlink_state(s, PF_ENTER_LOCKED);
2141 		killed++;
2142 
2143 		if (psk->psk_kill_match)
2144 			killed += pf_kill_matching_state(&match_key, dir);
2145 
2146 		goto relock_DIOCKILLSTATES;
2147 	}
2148 	PF_HASHROW_UNLOCK(ih);
2149 
2150 	return (killed);
2151 }
2152 
2153 static int
2154 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2155 {
2156 	int			 error = 0;
2157 	PF_RULES_RLOCK_TRACKER;
2158 
2159 #define	ERROUT_IOCTL(target, x)					\
2160     do {								\
2161 	    error = (x);						\
2162 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2163 	    goto target;						\
2164     } while (0)
2165 
2166 
2167 	/* XXX keep in sync with switch() below */
2168 	if (securelevel_gt(td->td_ucred, 2))
2169 		switch (cmd) {
2170 		case DIOCGETRULES:
2171 		case DIOCGETRULE:
2172 		case DIOCGETRULENV:
2173 		case DIOCGETADDRS:
2174 		case DIOCGETADDR:
2175 		case DIOCGETSTATE:
2176 		case DIOCGETSTATENV:
2177 		case DIOCSETSTATUSIF:
2178 		case DIOCGETSTATUS:
2179 		case DIOCGETSTATUSNV:
2180 		case DIOCCLRSTATUS:
2181 		case DIOCNATLOOK:
2182 		case DIOCSETDEBUG:
2183 		case DIOCGETSTATES:
2184 		case DIOCGETSTATESV2:
2185 		case DIOCGETTIMEOUT:
2186 		case DIOCCLRRULECTRS:
2187 		case DIOCGETLIMIT:
2188 		case DIOCGETALTQSV0:
2189 		case DIOCGETALTQSV1:
2190 		case DIOCGETALTQV0:
2191 		case DIOCGETALTQV1:
2192 		case DIOCGETQSTATSV0:
2193 		case DIOCGETQSTATSV1:
2194 		case DIOCGETRULESETS:
2195 		case DIOCGETRULESET:
2196 		case DIOCRGETTABLES:
2197 		case DIOCRGETTSTATS:
2198 		case DIOCRCLRTSTATS:
2199 		case DIOCRCLRADDRS:
2200 		case DIOCRADDADDRS:
2201 		case DIOCRDELADDRS:
2202 		case DIOCRSETADDRS:
2203 		case DIOCRGETADDRS:
2204 		case DIOCRGETASTATS:
2205 		case DIOCRCLRASTATS:
2206 		case DIOCRTSTADDRS:
2207 		case DIOCOSFPGET:
2208 		case DIOCGETSRCNODES:
2209 		case DIOCCLRSRCNODES:
2210 		case DIOCGETSYNCOOKIES:
2211 		case DIOCIGETIFACES:
2212 		case DIOCGIFSPEEDV0:
2213 		case DIOCGIFSPEEDV1:
2214 		case DIOCSETIFFLAG:
2215 		case DIOCCLRIFFLAG:
2216 			break;
2217 		case DIOCRCLRTABLES:
2218 		case DIOCRADDTABLES:
2219 		case DIOCRDELTABLES:
2220 		case DIOCRSETTFLAGS:
2221 			if (((struct pfioc_table *)addr)->pfrio_flags &
2222 			    PFR_FLAG_DUMMY)
2223 				break; /* dummy operation ok */
2224 			return (EPERM);
2225 		default:
2226 			return (EPERM);
2227 		}
2228 
2229 	if (!(flags & FWRITE))
2230 		switch (cmd) {
2231 		case DIOCGETRULES:
2232 		case DIOCGETADDRS:
2233 		case DIOCGETADDR:
2234 		case DIOCGETSTATE:
2235 		case DIOCGETSTATENV:
2236 		case DIOCGETSTATUS:
2237 		case DIOCGETSTATUSNV:
2238 		case DIOCGETSTATES:
2239 		case DIOCGETSTATESV2:
2240 		case DIOCGETTIMEOUT:
2241 		case DIOCGETLIMIT:
2242 		case DIOCGETALTQSV0:
2243 		case DIOCGETALTQSV1:
2244 		case DIOCGETALTQV0:
2245 		case DIOCGETALTQV1:
2246 		case DIOCGETQSTATSV0:
2247 		case DIOCGETQSTATSV1:
2248 		case DIOCGETRULESETS:
2249 		case DIOCGETRULESET:
2250 		case DIOCNATLOOK:
2251 		case DIOCRGETTABLES:
2252 		case DIOCRGETTSTATS:
2253 		case DIOCRGETADDRS:
2254 		case DIOCRGETASTATS:
2255 		case DIOCRTSTADDRS:
2256 		case DIOCOSFPGET:
2257 		case DIOCGETSRCNODES:
2258 		case DIOCGETSYNCOOKIES:
2259 		case DIOCIGETIFACES:
2260 		case DIOCGIFSPEEDV1:
2261 		case DIOCGIFSPEEDV0:
2262 		case DIOCGETRULENV:
2263 			break;
2264 		case DIOCRCLRTABLES:
2265 		case DIOCRADDTABLES:
2266 		case DIOCRDELTABLES:
2267 		case DIOCRCLRTSTATS:
2268 		case DIOCRCLRADDRS:
2269 		case DIOCRADDADDRS:
2270 		case DIOCRDELADDRS:
2271 		case DIOCRSETADDRS:
2272 		case DIOCRSETTFLAGS:
2273 			if (((struct pfioc_table *)addr)->pfrio_flags &
2274 			    PFR_FLAG_DUMMY) {
2275 				flags |= FWRITE; /* need write lock for dummy */
2276 				break; /* dummy operation ok */
2277 			}
2278 			return (EACCES);
2279 		case DIOCGETRULE:
2280 			if (((struct pfioc_rule *)addr)->action ==
2281 			    PF_GET_CLR_CNTR)
2282 				return (EACCES);
2283 			break;
2284 		default:
2285 			return (EACCES);
2286 		}
2287 
2288 	CURVNET_SET(TD_TO_VNET(td));
2289 
2290 	switch (cmd) {
2291 	case DIOCSTART:
2292 		sx_xlock(&pf_ioctl_lock);
2293 		if (V_pf_status.running)
2294 			error = EEXIST;
2295 		else {
2296 			int cpu;
2297 
2298 			hook_pf();
2299 			V_pf_status.running = 1;
2300 			V_pf_status.since = time_second;
2301 
2302 			CPU_FOREACH(cpu)
2303 				V_pf_stateid[cpu] = time_second;
2304 
2305 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
2306 		}
2307 		break;
2308 
2309 	case DIOCSTOP:
2310 		sx_xlock(&pf_ioctl_lock);
2311 		if (!V_pf_status.running)
2312 			error = ENOENT;
2313 		else {
2314 			V_pf_status.running = 0;
2315 			dehook_pf();
2316 			V_pf_status.since = time_second;
2317 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
2318 		}
2319 		break;
2320 
2321 	case DIOCADDRULENV: {
2322 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
2323 		nvlist_t	*nvl = NULL;
2324 		void		*nvlpacked = NULL;
2325 		struct pf_krule	*rule = NULL;
2326 		const char	*anchor = "", *anchor_call = "";
2327 		uint32_t	 ticket = 0, pool_ticket = 0;
2328 
2329 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
2330 
2331 		if (nv->len > pf_ioctl_maxcount)
2332 			ERROUT(ENOMEM);
2333 
2334 		nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
2335 		error = copyin(nv->data, nvlpacked, nv->len);
2336 		if (error)
2337 			ERROUT(error);
2338 
2339 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2340 		if (nvl == NULL)
2341 			ERROUT(EBADMSG);
2342 
2343 		if (! nvlist_exists_number(nvl, "ticket"))
2344 			ERROUT(EINVAL);
2345 		ticket = nvlist_get_number(nvl, "ticket");
2346 
2347 		if (! nvlist_exists_number(nvl, "pool_ticket"))
2348 			ERROUT(EINVAL);
2349 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
2350 
2351 		if (! nvlist_exists_nvlist(nvl, "rule"))
2352 			ERROUT(EINVAL);
2353 
2354 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK | M_ZERO);
2355 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
2356 		    rule);
2357 		if (error)
2358 			ERROUT(error);
2359 
2360 		if (nvlist_exists_string(nvl, "anchor"))
2361 			anchor = nvlist_get_string(nvl, "anchor");
2362 		if (nvlist_exists_string(nvl, "anchor_call"))
2363 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2364 
2365 		if ((error = nvlist_error(nvl)))
2366 			ERROUT(error);
2367 
2368 		/* Frees rule on error */
2369 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
2370 		    anchor_call, td);
2371 
2372 		nvlist_destroy(nvl);
2373 		free(nvlpacked, M_TEMP);
2374 		break;
2375 #undef ERROUT
2376 DIOCADDRULENV_error:
2377 		pf_krule_free(rule);
2378 		nvlist_destroy(nvl);
2379 		free(nvlpacked, M_TEMP);
2380 
2381 		break;
2382 	}
2383 	case DIOCADDRULE: {
2384 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2385 		struct pf_krule		*rule;
2386 
2387 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK | M_ZERO);
2388 		error = pf_rule_to_krule(&pr->rule, rule);
2389 		if (error != 0) {
2390 			free(rule, M_PFRULE);
2391 			break;
2392 		}
2393 
2394 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2395 
2396 		/* Frees rule on error */
2397 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
2398 		    pr->anchor, pr->anchor_call, td);
2399 		break;
2400 	}
2401 
2402 	case DIOCGETRULES: {
2403 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2404 		struct pf_kruleset	*ruleset;
2405 		struct pf_krule		*tail;
2406 		int			 rs_num;
2407 
2408 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2409 
2410 		PF_RULES_WLOCK();
2411 		ruleset = pf_find_kruleset(pr->anchor);
2412 		if (ruleset == NULL) {
2413 			PF_RULES_WUNLOCK();
2414 			error = EINVAL;
2415 			break;
2416 		}
2417 		rs_num = pf_get_ruleset_number(pr->rule.action);
2418 		if (rs_num >= PF_RULESET_MAX) {
2419 			PF_RULES_WUNLOCK();
2420 			error = EINVAL;
2421 			break;
2422 		}
2423 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
2424 		    pf_krulequeue);
2425 		if (tail)
2426 			pr->nr = tail->nr + 1;
2427 		else
2428 			pr->nr = 0;
2429 		pr->ticket = ruleset->rules[rs_num].active.ticket;
2430 		PF_RULES_WUNLOCK();
2431 		break;
2432 	}
2433 
2434 	case DIOCGETRULE: {
2435 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
2436 		struct pf_kruleset	*ruleset;
2437 		struct pf_krule		*rule;
2438 		int			 rs_num;
2439 
2440 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
2441 
2442 		PF_RULES_WLOCK();
2443 		ruleset = pf_find_kruleset(pr->anchor);
2444 		if (ruleset == NULL) {
2445 			PF_RULES_WUNLOCK();
2446 			error = EINVAL;
2447 			break;
2448 		}
2449 		rs_num = pf_get_ruleset_number(pr->rule.action);
2450 		if (rs_num >= PF_RULESET_MAX) {
2451 			PF_RULES_WUNLOCK();
2452 			error = EINVAL;
2453 			break;
2454 		}
2455 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
2456 			PF_RULES_WUNLOCK();
2457 			error = EBUSY;
2458 			break;
2459 		}
2460 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2461 		while ((rule != NULL) && (rule->nr != pr->nr))
2462 			rule = TAILQ_NEXT(rule, entries);
2463 		if (rule == NULL) {
2464 			PF_RULES_WUNLOCK();
2465 			error = EBUSY;
2466 			break;
2467 		}
2468 
2469 		pf_krule_to_rule(rule, &pr->rule);
2470 
2471 		if (pf_kanchor_copyout(ruleset, rule, pr)) {
2472 			PF_RULES_WUNLOCK();
2473 			error = EBUSY;
2474 			break;
2475 		}
2476 		pf_addr_copyout(&pr->rule.src.addr);
2477 		pf_addr_copyout(&pr->rule.dst.addr);
2478 
2479 		if (pr->action == PF_GET_CLR_CNTR) {
2480 			pf_counter_u64_zero(&rule->evaluations);
2481 			for (int i = 0; i < 2; i++) {
2482 				pf_counter_u64_zero(&rule->packets[i]);
2483 				pf_counter_u64_zero(&rule->bytes[i]);
2484 			}
2485 			counter_u64_zero(rule->states_tot);
2486 		}
2487 		PF_RULES_WUNLOCK();
2488 		break;
2489 	}
2490 
2491 	case DIOCGETRULENV: {
2492 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2493 		nvlist_t		*nvrule = NULL;
2494 		nvlist_t		*nvl = NULL;
2495 		struct pf_kruleset	*ruleset;
2496 		struct pf_krule		*rule;
2497 		void			*nvlpacked = NULL;
2498 		int			 rs_num, nr;
2499 		bool			 clear_counter = false;
2500 
2501 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
2502 
2503 		if (nv->len > pf_ioctl_maxcount)
2504 			ERROUT(ENOMEM);
2505 
2506 		/* Copy the request in */
2507 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2508 		if (nvlpacked == NULL)
2509 			ERROUT(ENOMEM);
2510 
2511 		error = copyin(nv->data, nvlpacked, nv->len);
2512 		if (error)
2513 			ERROUT(error);
2514 
2515 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2516 		if (nvl == NULL)
2517 			ERROUT(EBADMSG);
2518 
2519 		if (! nvlist_exists_string(nvl, "anchor"))
2520 			ERROUT(EBADMSG);
2521 		if (! nvlist_exists_number(nvl, "ruleset"))
2522 			ERROUT(EBADMSG);
2523 		if (! nvlist_exists_number(nvl, "ticket"))
2524 			ERROUT(EBADMSG);
2525 		if (! nvlist_exists_number(nvl, "nr"))
2526 			ERROUT(EBADMSG);
2527 
2528 		if (nvlist_exists_bool(nvl, "clear_counter"))
2529 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
2530 
2531 		if (clear_counter && !(flags & FWRITE))
2532 			ERROUT(EACCES);
2533 
2534 		nr = nvlist_get_number(nvl, "nr");
2535 
2536 		PF_RULES_WLOCK();
2537 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
2538 		if (ruleset == NULL) {
2539 			PF_RULES_WUNLOCK();
2540 			ERROUT(ENOENT);
2541 		}
2542 
2543 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
2544 		if (rs_num >= PF_RULESET_MAX) {
2545 			PF_RULES_WUNLOCK();
2546 			ERROUT(EINVAL);
2547 		}
2548 
2549 		if (nvlist_get_number(nvl, "ticket") !=
2550 		    ruleset->rules[rs_num].active.ticket) {
2551 			PF_RULES_WUNLOCK();
2552 			ERROUT(EBUSY);
2553 		}
2554 
2555 		if ((error = nvlist_error(nvl))) {
2556 			PF_RULES_WUNLOCK();
2557 			ERROUT(error);
2558 		}
2559 
2560 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
2561 		while ((rule != NULL) && (rule->nr != nr))
2562 			rule = TAILQ_NEXT(rule, entries);
2563 		if (rule == NULL) {
2564 			PF_RULES_WUNLOCK();
2565 			ERROUT(EBUSY);
2566 		}
2567 
2568 		nvrule = pf_krule_to_nvrule(rule);
2569 
2570 		nvlist_destroy(nvl);
2571 		nvl = nvlist_create(0);
2572 		if (nvl == NULL) {
2573 			PF_RULES_WUNLOCK();
2574 			ERROUT(ENOMEM);
2575 		}
2576 		nvlist_add_number(nvl, "nr", nr);
2577 		nvlist_add_nvlist(nvl, "rule", nvrule);
2578 		nvlist_destroy(nvrule);
2579 		nvrule = NULL;
2580 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
2581 			PF_RULES_WUNLOCK();
2582 			ERROUT(EBUSY);
2583 		}
2584 
2585 		free(nvlpacked, M_NVLIST);
2586 		nvlpacked = nvlist_pack(nvl, &nv->len);
2587 		if (nvlpacked == NULL) {
2588 			PF_RULES_WUNLOCK();
2589 			ERROUT(ENOMEM);
2590 		}
2591 
2592 		if (nv->size == 0) {
2593 			PF_RULES_WUNLOCK();
2594 			ERROUT(0);
2595 		}
2596 		else if (nv->size < nv->len) {
2597 			PF_RULES_WUNLOCK();
2598 			ERROUT(ENOSPC);
2599 		}
2600 
2601 		if (clear_counter) {
2602 			pf_counter_u64_zero(&rule->evaluations);
2603 			for (int i = 0; i < 2; i++) {
2604 				pf_counter_u64_zero(&rule->packets[i]);
2605 				pf_counter_u64_zero(&rule->bytes[i]);
2606 			}
2607 			counter_u64_zero(rule->states_tot);
2608 		}
2609 		PF_RULES_WUNLOCK();
2610 
2611 		error = copyout(nvlpacked, nv->data, nv->len);
2612 
2613 #undef ERROUT
2614 DIOCGETRULENV_error:
2615 		free(nvlpacked, M_NVLIST);
2616 		nvlist_destroy(nvrule);
2617 		nvlist_destroy(nvl);
2618 
2619 		break;
2620 	}
2621 
2622 	case DIOCCHANGERULE: {
2623 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
2624 		struct pf_kruleset	*ruleset;
2625 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
2626 		struct pfi_kkif		*kif = NULL;
2627 		struct pf_kpooladdr	*pa;
2628 		u_int32_t		 nr = 0;
2629 		int			 rs_num;
2630 
2631 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
2632 
2633 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
2634 		    pcr->action > PF_CHANGE_GET_TICKET) {
2635 			error = EINVAL;
2636 			break;
2637 		}
2638 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
2639 			error = EINVAL;
2640 			break;
2641 		}
2642 
2643 		if (pcr->action != PF_CHANGE_REMOVE) {
2644 			newrule = malloc(sizeof(*newrule), M_PFRULE, M_WAITOK | M_ZERO);
2645 			error = pf_rule_to_krule(&pcr->rule, newrule);
2646 			if (error != 0) {
2647 				free(newrule, M_PFRULE);
2648 				break;
2649 			}
2650 
2651 			if (newrule->ifname[0])
2652 				kif = pf_kkif_create(M_WAITOK);
2653 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
2654 			for (int i = 0; i < 2; i++) {
2655 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
2656 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
2657 			}
2658 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
2659 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
2660 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
2661 			newrule->cuid = td->td_ucred->cr_ruid;
2662 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
2663 			TAILQ_INIT(&newrule->rpool.list);
2664 		}
2665 #define	ERROUT(x)	{ error = (x); goto DIOCCHANGERULE_error; }
2666 
2667 		PF_RULES_WLOCK();
2668 #ifdef PF_WANT_32_TO_64_COUNTER
2669 		if (newrule != NULL) {
2670 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
2671 			newrule->allrulelinked = true;
2672 			V_pf_allrulecount++;
2673 		}
2674 #endif
2675 
2676 		if (!(pcr->action == PF_CHANGE_REMOVE ||
2677 		    pcr->action == PF_CHANGE_GET_TICKET) &&
2678 		    pcr->pool_ticket != V_ticket_pabuf)
2679 			ERROUT(EBUSY);
2680 
2681 		ruleset = pf_find_kruleset(pcr->anchor);
2682 		if (ruleset == NULL)
2683 			ERROUT(EINVAL);
2684 
2685 		rs_num = pf_get_ruleset_number(pcr->rule.action);
2686 		if (rs_num >= PF_RULESET_MAX)
2687 			ERROUT(EINVAL);
2688 
2689 		if (pcr->action == PF_CHANGE_GET_TICKET) {
2690 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
2691 			ERROUT(0);
2692 		} else if (pcr->ticket !=
2693 			    ruleset->rules[rs_num].active.ticket)
2694 				ERROUT(EINVAL);
2695 
2696 		if (pcr->action != PF_CHANGE_REMOVE) {
2697 			if (newrule->ifname[0]) {
2698 				newrule->kif = pfi_kkif_attach(kif,
2699 				    newrule->ifname);
2700 				kif = NULL;
2701 				pfi_kkif_ref(newrule->kif);
2702 			} else
2703 				newrule->kif = NULL;
2704 
2705 			if (newrule->rtableid > 0 &&
2706 			    newrule->rtableid >= rt_numfibs)
2707 				error = EBUSY;
2708 
2709 #ifdef ALTQ
2710 			/* set queue IDs */
2711 			if (newrule->qname[0] != 0) {
2712 				if ((newrule->qid =
2713 				    pf_qname2qid(newrule->qname)) == 0)
2714 					error = EBUSY;
2715 				else if (newrule->pqname[0] != 0) {
2716 					if ((newrule->pqid =
2717 					    pf_qname2qid(newrule->pqname)) == 0)
2718 						error = EBUSY;
2719 				} else
2720 					newrule->pqid = newrule->qid;
2721 			}
2722 #endif /* ALTQ */
2723 			if (newrule->tagname[0])
2724 				if ((newrule->tag =
2725 				    pf_tagname2tag(newrule->tagname)) == 0)
2726 					error = EBUSY;
2727 			if (newrule->match_tagname[0])
2728 				if ((newrule->match_tag = pf_tagname2tag(
2729 				    newrule->match_tagname)) == 0)
2730 					error = EBUSY;
2731 			if (newrule->rt && !newrule->direction)
2732 				error = EINVAL;
2733 			if (!newrule->log)
2734 				newrule->logif = 0;
2735 			if (newrule->logif >= PFLOGIFS_MAX)
2736 				error = EINVAL;
2737 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
2738 				error = ENOMEM;
2739 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
2740 				error = ENOMEM;
2741 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
2742 				error = EINVAL;
2743 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
2744 				if (pa->addr.type == PF_ADDR_TABLE) {
2745 					pa->addr.p.tbl =
2746 					    pfr_attach_table(ruleset,
2747 					    pa->addr.v.tblname);
2748 					if (pa->addr.p.tbl == NULL)
2749 						error = ENOMEM;
2750 				}
2751 
2752 			newrule->overload_tbl = NULL;
2753 			if (newrule->overload_tblname[0]) {
2754 				if ((newrule->overload_tbl = pfr_attach_table(
2755 				    ruleset, newrule->overload_tblname)) ==
2756 				    NULL)
2757 					error = EINVAL;
2758 				else
2759 					newrule->overload_tbl->pfrkt_flags |=
2760 					    PFR_TFLAG_ACTIVE;
2761 			}
2762 
2763 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
2764 			if (((((newrule->action == PF_NAT) ||
2765 			    (newrule->action == PF_RDR) ||
2766 			    (newrule->action == PF_BINAT) ||
2767 			    (newrule->rt > PF_NOPFROUTE)) &&
2768 			    !newrule->anchor)) &&
2769 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
2770 				error = EINVAL;
2771 
2772 			if (error) {
2773 				pf_free_rule(newrule);
2774 				PF_RULES_WUNLOCK();
2775 				break;
2776 			}
2777 
2778 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
2779 		}
2780 		pf_empty_kpool(&V_pf_pabuf);
2781 
2782 		if (pcr->action == PF_CHANGE_ADD_HEAD)
2783 			oldrule = TAILQ_FIRST(
2784 			    ruleset->rules[rs_num].active.ptr);
2785 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
2786 			oldrule = TAILQ_LAST(
2787 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
2788 		else {
2789 			oldrule = TAILQ_FIRST(
2790 			    ruleset->rules[rs_num].active.ptr);
2791 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
2792 				oldrule = TAILQ_NEXT(oldrule, entries);
2793 			if (oldrule == NULL) {
2794 				if (newrule != NULL)
2795 					pf_free_rule(newrule);
2796 				PF_RULES_WUNLOCK();
2797 				error = EINVAL;
2798 				break;
2799 			}
2800 		}
2801 
2802 		if (pcr->action == PF_CHANGE_REMOVE) {
2803 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
2804 			    oldrule);
2805 			ruleset->rules[rs_num].active.rcount--;
2806 		} else {
2807 			if (oldrule == NULL)
2808 				TAILQ_INSERT_TAIL(
2809 				    ruleset->rules[rs_num].active.ptr,
2810 				    newrule, entries);
2811 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
2812 			    pcr->action == PF_CHANGE_ADD_BEFORE)
2813 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
2814 			else
2815 				TAILQ_INSERT_AFTER(
2816 				    ruleset->rules[rs_num].active.ptr,
2817 				    oldrule, newrule, entries);
2818 			ruleset->rules[rs_num].active.rcount++;
2819 		}
2820 
2821 		nr = 0;
2822 		TAILQ_FOREACH(oldrule,
2823 		    ruleset->rules[rs_num].active.ptr, entries)
2824 			oldrule->nr = nr++;
2825 
2826 		ruleset->rules[rs_num].active.ticket++;
2827 
2828 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
2829 		pf_remove_if_empty_kruleset(ruleset);
2830 
2831 		PF_RULES_WUNLOCK();
2832 		break;
2833 
2834 #undef ERROUT
2835 DIOCCHANGERULE_error:
2836 		PF_RULES_WUNLOCK();
2837 		pf_krule_free(newrule);
2838 		pf_kkif_free(kif);
2839 		break;
2840 	}
2841 
2842 	case DIOCCLRSTATES: {
2843 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
2844 		struct pf_kstate_kill	 kill;
2845 
2846 		error = pf_state_kill_to_kstate_kill(psk, &kill);
2847 		if (error)
2848 			break;
2849 
2850 		psk->psk_killed = pf_clear_states(&kill);
2851 		break;
2852 	}
2853 
2854 	case DIOCCLRSTATESNV: {
2855 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
2856 		break;
2857 	}
2858 
2859 	case DIOCKILLSTATES: {
2860 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
2861 		struct pf_kstate_kill	 kill;
2862 
2863 		error = pf_state_kill_to_kstate_kill(psk, &kill);
2864 		if (error)
2865 			break;
2866 
2867 		psk->psk_killed = 0;
2868 		pf_killstates(&kill, &psk->psk_killed);
2869 		break;
2870 	}
2871 
2872 	case DIOCKILLSTATESNV: {
2873 		error = pf_killstates_nv((struct pfioc_nv *)addr);
2874 		break;
2875 	}
2876 
2877 	case DIOCADDSTATE: {
2878 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2879 		struct pfsync_state	*sp = &ps->state;
2880 
2881 		if (sp->timeout >= PFTM_MAX) {
2882 			error = EINVAL;
2883 			break;
2884 		}
2885 		if (V_pfsync_state_import_ptr != NULL) {
2886 			PF_RULES_RLOCK();
2887 			error = V_pfsync_state_import_ptr(sp, PFSYNC_SI_IOCTL);
2888 			PF_RULES_RUNLOCK();
2889 		} else
2890 			error = EOPNOTSUPP;
2891 		break;
2892 	}
2893 
2894 	case DIOCGETSTATE: {
2895 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
2896 		struct pf_kstate	*s;
2897 
2898 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
2899 		if (s == NULL) {
2900 			error = ENOENT;
2901 			break;
2902 		}
2903 
2904 		pfsync_state_export(&ps->state, s);
2905 		PF_STATE_UNLOCK(s);
2906 		break;
2907 	}
2908 
2909 	case DIOCGETSTATENV: {
2910 		error = pf_getstate((struct pfioc_nv *)addr);
2911 		break;
2912 	}
2913 
2914 	case DIOCGETSTATES: {
2915 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2916 		struct pf_kstate	*s;
2917 		struct pfsync_state	*pstore, *p;
2918 		int			 i, nr;
2919 		size_t			 slice_count = 16, count;
2920 		void			*out;
2921 
2922 		if (ps->ps_len <= 0) {
2923 			nr = uma_zone_get_cur(V_pf_state_z);
2924 			ps->ps_len = sizeof(struct pfsync_state) * nr;
2925 			break;
2926 		}
2927 
2928 		out = ps->ps_states;
2929 		pstore = mallocarray(slice_count,
2930 		    sizeof(struct pfsync_state), M_TEMP, M_WAITOK | M_ZERO);
2931 		nr = 0;
2932 
2933 		for (i = 0; i <= pf_hashmask; i++) {
2934 			struct pf_idhash *ih = &V_pf_idhash[i];
2935 
2936 DIOCGETSTATES_retry:
2937 			p = pstore;
2938 
2939 			if (LIST_EMPTY(&ih->states))
2940 				continue;
2941 
2942 			PF_HASHROW_LOCK(ih);
2943 			count = 0;
2944 			LIST_FOREACH(s, &ih->states, entry) {
2945 				if (s->timeout == PFTM_UNLINKED)
2946 					continue;
2947 				count++;
2948 			}
2949 
2950 			if (count > slice_count) {
2951 				PF_HASHROW_UNLOCK(ih);
2952 				free(pstore, M_TEMP);
2953 				slice_count = count * 2;
2954 				pstore = mallocarray(slice_count,
2955 				    sizeof(struct pfsync_state), M_TEMP,
2956 				    M_WAITOK | M_ZERO);
2957 				goto DIOCGETSTATES_retry;
2958 			}
2959 
2960 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
2961 				PF_HASHROW_UNLOCK(ih);
2962 				goto DIOCGETSTATES_full;
2963 			}
2964 
2965 			LIST_FOREACH(s, &ih->states, entry) {
2966 				if (s->timeout == PFTM_UNLINKED)
2967 					continue;
2968 
2969 				pfsync_state_export(p, s);
2970 				p++;
2971 				nr++;
2972 			}
2973 			PF_HASHROW_UNLOCK(ih);
2974 			error = copyout(pstore, out,
2975 			    sizeof(struct pfsync_state) * count);
2976 			if (error)
2977 				break;
2978 			out = ps->ps_states + nr;
2979 		}
2980 DIOCGETSTATES_full:
2981 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2982 		free(pstore, M_TEMP);
2983 
2984 		break;
2985 	}
2986 
2987 	case DIOCGETSTATESV2: {
2988 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
2989 		struct pf_kstate	*s;
2990 		struct pf_state_export	*pstore, *p;
2991 		int i, nr;
2992 		size_t slice_count = 16, count;
2993 		void *out;
2994 
2995 		if (ps->ps_req_version > PF_STATE_VERSION) {
2996 			error = ENOTSUP;
2997 			break;
2998 		}
2999 
3000 		if (ps->ps_len <= 0) {
3001 			nr = uma_zone_get_cur(V_pf_state_z);
3002 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3003 			break;
3004 		}
3005 
3006 		out = ps->ps_states;
3007 		pstore = mallocarray(slice_count,
3008 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3009 		nr = 0;
3010 
3011 		for (i = 0; i <= pf_hashmask; i++) {
3012 			struct pf_idhash *ih = &V_pf_idhash[i];
3013 
3014 DIOCGETSTATESV2_retry:
3015 			p = pstore;
3016 
3017 			if (LIST_EMPTY(&ih->states))
3018 				continue;
3019 
3020 			PF_HASHROW_LOCK(ih);
3021 			count = 0;
3022 			LIST_FOREACH(s, &ih->states, entry) {
3023 				if (s->timeout == PFTM_UNLINKED)
3024 					continue;
3025 				count++;
3026 			}
3027 
3028 			if (count > slice_count) {
3029 				PF_HASHROW_UNLOCK(ih);
3030 				free(pstore, M_TEMP);
3031 				slice_count = count * 2;
3032 				pstore = mallocarray(slice_count,
3033 				    sizeof(struct pf_state_export), M_TEMP,
3034 				    M_WAITOK | M_ZERO);
3035 				goto DIOCGETSTATESV2_retry;
3036 			}
3037 
3038 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3039 				PF_HASHROW_UNLOCK(ih);
3040 				goto DIOCGETSTATESV2_full;
3041 			}
3042 
3043 			LIST_FOREACH(s, &ih->states, entry) {
3044 				if (s->timeout == PFTM_UNLINKED)
3045 					continue;
3046 
3047 				pf_state_export(p, s);
3048 				p++;
3049 				nr++;
3050 			}
3051 			PF_HASHROW_UNLOCK(ih);
3052 			error = copyout(pstore, out,
3053 			    sizeof(struct pf_state_export) * count);
3054 			if (error)
3055 				break;
3056 			out = ps->ps_states + nr;
3057 		}
3058 DIOCGETSTATESV2_full:
3059 		ps->ps_len = nr * sizeof(struct pf_state_export);
3060 		free(pstore, M_TEMP);
3061 
3062 		break;
3063 	}
3064 
3065 	case DIOCGETSTATUS: {
3066 		struct pf_status *s = (struct pf_status *)addr;
3067 
3068 		PF_RULES_RLOCK();
3069 		s->running = V_pf_status.running;
3070 		s->since   = V_pf_status.since;
3071 		s->debug   = V_pf_status.debug;
3072 		s->hostid  = V_pf_status.hostid;
3073 		s->states  = V_pf_status.states;
3074 		s->src_nodes = V_pf_status.src_nodes;
3075 
3076 		for (int i = 0; i < PFRES_MAX; i++)
3077 			s->counters[i] =
3078 			    counter_u64_fetch(V_pf_status.counters[i]);
3079 		for (int i = 0; i < LCNT_MAX; i++)
3080 			s->lcounters[i] =
3081 			    counter_u64_fetch(V_pf_status.lcounters[i]);
3082 		for (int i = 0; i < FCNT_MAX; i++)
3083 			s->fcounters[i] =
3084 			    pf_counter_u64_fetch(&V_pf_status.fcounters[i]);
3085 		for (int i = 0; i < SCNT_MAX; i++)
3086 			s->scounters[i] =
3087 			    counter_u64_fetch(V_pf_status.scounters[i]);
3088 
3089 		bcopy(V_pf_status.ifname, s->ifname, IFNAMSIZ);
3090 		bcopy(V_pf_status.pf_chksum, s->pf_chksum,
3091 		    PF_MD5_DIGEST_LENGTH);
3092 
3093 		pfi_update_status(s->ifname, s);
3094 		PF_RULES_RUNLOCK();
3095 		break;
3096 	}
3097 
3098 	case DIOCGETSTATUSNV: {
3099 		error = pf_getstatus((struct pfioc_nv *)addr);
3100 		break;
3101 	}
3102 
3103 	case DIOCSETSTATUSIF: {
3104 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3105 
3106 		if (pi->ifname[0] == 0) {
3107 			bzero(V_pf_status.ifname, IFNAMSIZ);
3108 			break;
3109 		}
3110 		PF_RULES_WLOCK();
3111 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3112 		PF_RULES_WUNLOCK();
3113 		break;
3114 	}
3115 
3116 	case DIOCCLRSTATUS: {
3117 		PF_RULES_WLOCK();
3118 		for (int i = 0; i < PFRES_MAX; i++)
3119 			counter_u64_zero(V_pf_status.counters[i]);
3120 		for (int i = 0; i < FCNT_MAX; i++)
3121 			pf_counter_u64_zero(&V_pf_status.fcounters[i]);
3122 		for (int i = 0; i < SCNT_MAX; i++)
3123 			counter_u64_zero(V_pf_status.scounters[i]);
3124 		for (int i = 0; i < KLCNT_MAX; i++)
3125 			counter_u64_zero(V_pf_status.lcounters[i]);
3126 		V_pf_status.since = time_second;
3127 		if (*V_pf_status.ifname)
3128 			pfi_update_status(V_pf_status.ifname, NULL);
3129 		PF_RULES_WUNLOCK();
3130 		break;
3131 	}
3132 
3133 	case DIOCNATLOOK: {
3134 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3135 		struct pf_state_key	*sk;
3136 		struct pf_kstate	*state;
3137 		struct pf_state_key_cmp	 key;
3138 		int			 m = 0, direction = pnl->direction;
3139 		int			 sidx, didx;
3140 
3141 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3142 		sidx = (direction == PF_IN) ? 1 : 0;
3143 		didx = (direction == PF_IN) ? 0 : 1;
3144 
3145 		if (!pnl->proto ||
3146 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3147 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3148 		    ((pnl->proto == IPPROTO_TCP ||
3149 		    pnl->proto == IPPROTO_UDP) &&
3150 		    (!pnl->dport || !pnl->sport)))
3151 			error = EINVAL;
3152 		else {
3153 			bzero(&key, sizeof(key));
3154 			key.af = pnl->af;
3155 			key.proto = pnl->proto;
3156 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3157 			key.port[sidx] = pnl->sport;
3158 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3159 			key.port[didx] = pnl->dport;
3160 
3161 			state = pf_find_state_all(&key, direction, &m);
3162 
3163 			if (m > 1)
3164 				error = E2BIG;	/* more than one state */
3165 			else if (state != NULL) {
3166 				/* XXXGL: not locked read */
3167 				sk = state->key[sidx];
3168 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
3169 				pnl->rsport = sk->port[sidx];
3170 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
3171 				pnl->rdport = sk->port[didx];
3172 			} else
3173 				error = ENOENT;
3174 		}
3175 		break;
3176 	}
3177 
3178 	case DIOCSETTIMEOUT: {
3179 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3180 		int		 old;
3181 
3182 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
3183 		    pt->seconds < 0) {
3184 			error = EINVAL;
3185 			break;
3186 		}
3187 		PF_RULES_WLOCK();
3188 		old = V_pf_default_rule.timeout[pt->timeout];
3189 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
3190 			pt->seconds = 1;
3191 		V_pf_default_rule.timeout[pt->timeout] = pt->seconds;
3192 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
3193 			wakeup(pf_purge_thread);
3194 		pt->seconds = old;
3195 		PF_RULES_WUNLOCK();
3196 		break;
3197 	}
3198 
3199 	case DIOCGETTIMEOUT: {
3200 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
3201 
3202 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
3203 			error = EINVAL;
3204 			break;
3205 		}
3206 		PF_RULES_RLOCK();
3207 		pt->seconds = V_pf_default_rule.timeout[pt->timeout];
3208 		PF_RULES_RUNLOCK();
3209 		break;
3210 	}
3211 
3212 	case DIOCGETLIMIT: {
3213 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3214 
3215 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
3216 			error = EINVAL;
3217 			break;
3218 		}
3219 		PF_RULES_RLOCK();
3220 		pl->limit = V_pf_limits[pl->index].limit;
3221 		PF_RULES_RUNLOCK();
3222 		break;
3223 	}
3224 
3225 	case DIOCSETLIMIT: {
3226 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
3227 		int			 old_limit;
3228 
3229 		PF_RULES_WLOCK();
3230 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
3231 		    V_pf_limits[pl->index].zone == NULL) {
3232 			PF_RULES_WUNLOCK();
3233 			error = EINVAL;
3234 			break;
3235 		}
3236 		uma_zone_set_max(V_pf_limits[pl->index].zone, pl->limit);
3237 		old_limit = V_pf_limits[pl->index].limit;
3238 		V_pf_limits[pl->index].limit = pl->limit;
3239 		pl->limit = old_limit;
3240 		PF_RULES_WUNLOCK();
3241 		break;
3242 	}
3243 
3244 	case DIOCSETDEBUG: {
3245 		u_int32_t	*level = (u_int32_t *)addr;
3246 
3247 		PF_RULES_WLOCK();
3248 		V_pf_status.debug = *level;
3249 		PF_RULES_WUNLOCK();
3250 		break;
3251 	}
3252 
3253 	case DIOCCLRRULECTRS: {
3254 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
3255 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
3256 		struct pf_krule		*rule;
3257 
3258 		PF_RULES_WLOCK();
3259 		TAILQ_FOREACH(rule,
3260 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
3261 			pf_counter_u64_zero(&rule->evaluations);
3262 			for (int i = 0; i < 2; i++) {
3263 				pf_counter_u64_zero(&rule->packets[i]);
3264 				pf_counter_u64_zero(&rule->bytes[i]);
3265 			}
3266 		}
3267 		PF_RULES_WUNLOCK();
3268 		break;
3269 	}
3270 
3271 	case DIOCGIFSPEEDV0:
3272 	case DIOCGIFSPEEDV1: {
3273 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
3274 		struct pf_ifspeed_v1	ps;
3275 		struct ifnet		*ifp;
3276 
3277 		if (psp->ifname[0] == '\0') {
3278 			error = EINVAL;
3279 			break;
3280 		}
3281 
3282 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
3283 		if (error != 0)
3284 			break;
3285 		ifp = ifunit(ps.ifname);
3286 		if (ifp != NULL) {
3287 			psp->baudrate32 =
3288 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
3289 			if (cmd == DIOCGIFSPEEDV1)
3290 				psp->baudrate = ifp->if_baudrate;
3291 		} else {
3292 			error = EINVAL;
3293 		}
3294 		break;
3295 	}
3296 
3297 #ifdef ALTQ
3298 	case DIOCSTARTALTQ: {
3299 		struct pf_altq		*altq;
3300 
3301 		PF_RULES_WLOCK();
3302 		/* enable all altq interfaces on active list */
3303 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3304 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3305 				error = pf_enable_altq(altq);
3306 				if (error != 0)
3307 					break;
3308 			}
3309 		}
3310 		if (error == 0)
3311 			V_pf_altq_running = 1;
3312 		PF_RULES_WUNLOCK();
3313 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
3314 		break;
3315 	}
3316 
3317 	case DIOCSTOPALTQ: {
3318 		struct pf_altq		*altq;
3319 
3320 		PF_RULES_WLOCK();
3321 		/* disable all altq interfaces on active list */
3322 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
3323 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
3324 				error = pf_disable_altq(altq);
3325 				if (error != 0)
3326 					break;
3327 			}
3328 		}
3329 		if (error == 0)
3330 			V_pf_altq_running = 0;
3331 		PF_RULES_WUNLOCK();
3332 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
3333 		break;
3334 	}
3335 
3336 	case DIOCADDALTQV0:
3337 	case DIOCADDALTQV1: {
3338 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3339 		struct pf_altq		*altq, *a;
3340 		struct ifnet		*ifp;
3341 
3342 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
3343 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
3344 		if (error)
3345 			break;
3346 		altq->local_flags = 0;
3347 
3348 		PF_RULES_WLOCK();
3349 		if (pa->ticket != V_ticket_altqs_inactive) {
3350 			PF_RULES_WUNLOCK();
3351 			free(altq, M_PFALTQ);
3352 			error = EBUSY;
3353 			break;
3354 		}
3355 
3356 		/*
3357 		 * if this is for a queue, find the discipline and
3358 		 * copy the necessary fields
3359 		 */
3360 		if (altq->qname[0] != 0) {
3361 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
3362 				PF_RULES_WUNLOCK();
3363 				error = EBUSY;
3364 				free(altq, M_PFALTQ);
3365 				break;
3366 			}
3367 			altq->altq_disc = NULL;
3368 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
3369 				if (strncmp(a->ifname, altq->ifname,
3370 				    IFNAMSIZ) == 0) {
3371 					altq->altq_disc = a->altq_disc;
3372 					break;
3373 				}
3374 			}
3375 		}
3376 
3377 		if ((ifp = ifunit(altq->ifname)) == NULL)
3378 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
3379 		else
3380 			error = altq_add(ifp, altq);
3381 
3382 		if (error) {
3383 			PF_RULES_WUNLOCK();
3384 			free(altq, M_PFALTQ);
3385 			break;
3386 		}
3387 
3388 		if (altq->qname[0] != 0)
3389 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
3390 		else
3391 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
3392 		/* version error check done on import above */
3393 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
3394 		PF_RULES_WUNLOCK();
3395 		break;
3396 	}
3397 
3398 	case DIOCGETALTQSV0:
3399 	case DIOCGETALTQSV1: {
3400 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3401 		struct pf_altq		*altq;
3402 
3403 		PF_RULES_RLOCK();
3404 		pa->nr = 0;
3405 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
3406 			pa->nr++;
3407 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
3408 			pa->nr++;
3409 		pa->ticket = V_ticket_altqs_active;
3410 		PF_RULES_RUNLOCK();
3411 		break;
3412 	}
3413 
3414 	case DIOCGETALTQV0:
3415 	case DIOCGETALTQV1: {
3416 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
3417 		struct pf_altq		*altq;
3418 
3419 		PF_RULES_RLOCK();
3420 		if (pa->ticket != V_ticket_altqs_active) {
3421 			PF_RULES_RUNLOCK();
3422 			error = EBUSY;
3423 			break;
3424 		}
3425 		altq = pf_altq_get_nth_active(pa->nr);
3426 		if (altq == NULL) {
3427 			PF_RULES_RUNLOCK();
3428 			error = EBUSY;
3429 			break;
3430 		}
3431 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
3432 		PF_RULES_RUNLOCK();
3433 		break;
3434 	}
3435 
3436 	case DIOCCHANGEALTQV0:
3437 	case DIOCCHANGEALTQV1:
3438 		/* CHANGEALTQ not supported yet! */
3439 		error = ENODEV;
3440 		break;
3441 
3442 	case DIOCGETQSTATSV0:
3443 	case DIOCGETQSTATSV1: {
3444 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
3445 		struct pf_altq		*altq;
3446 		int			 nbytes;
3447 		u_int32_t		 version;
3448 
3449 		PF_RULES_RLOCK();
3450 		if (pq->ticket != V_ticket_altqs_active) {
3451 			PF_RULES_RUNLOCK();
3452 			error = EBUSY;
3453 			break;
3454 		}
3455 		nbytes = pq->nbytes;
3456 		altq = pf_altq_get_nth_active(pq->nr);
3457 		if (altq == NULL) {
3458 			PF_RULES_RUNLOCK();
3459 			error = EBUSY;
3460 			break;
3461 		}
3462 
3463 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
3464 			PF_RULES_RUNLOCK();
3465 			error = ENXIO;
3466 			break;
3467 		}
3468 		PF_RULES_RUNLOCK();
3469 		if (cmd == DIOCGETQSTATSV0)
3470 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
3471 		else
3472 			version = pq->version;
3473 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
3474 		if (error == 0) {
3475 			pq->scheduler = altq->scheduler;
3476 			pq->nbytes = nbytes;
3477 		}
3478 		break;
3479 	}
3480 #endif /* ALTQ */
3481 
3482 	case DIOCBEGINADDRS: {
3483 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3484 
3485 		PF_RULES_WLOCK();
3486 		pf_empty_kpool(&V_pf_pabuf);
3487 		pp->ticket = ++V_ticket_pabuf;
3488 		PF_RULES_WUNLOCK();
3489 		break;
3490 	}
3491 
3492 	case DIOCADDADDR: {
3493 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3494 		struct pf_kpooladdr	*pa;
3495 		struct pfi_kkif		*kif = NULL;
3496 
3497 #ifndef INET
3498 		if (pp->af == AF_INET) {
3499 			error = EAFNOSUPPORT;
3500 			break;
3501 		}
3502 #endif /* INET */
3503 #ifndef INET6
3504 		if (pp->af == AF_INET6) {
3505 			error = EAFNOSUPPORT;
3506 			break;
3507 		}
3508 #endif /* INET6 */
3509 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
3510 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
3511 		    pp->addr.addr.type != PF_ADDR_TABLE) {
3512 			error = EINVAL;
3513 			break;
3514 		}
3515 		if (pp->addr.addr.p.dyn != NULL) {
3516 			error = EINVAL;
3517 			break;
3518 		}
3519 		pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
3520 		error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
3521 		if (error != 0)
3522 			break;
3523 		if (pa->ifname[0])
3524 			kif = pf_kkif_create(M_WAITOK);
3525 		PF_RULES_WLOCK();
3526 		if (pp->ticket != V_ticket_pabuf) {
3527 			PF_RULES_WUNLOCK();
3528 			if (pa->ifname[0])
3529 				pf_kkif_free(kif);
3530 			free(pa, M_PFRULE);
3531 			error = EBUSY;
3532 			break;
3533 		}
3534 		if (pa->ifname[0]) {
3535 			pa->kif = pfi_kkif_attach(kif, pa->ifname);
3536 			kif = NULL;
3537 			pfi_kkif_ref(pa->kif);
3538 		} else
3539 			pa->kif = NULL;
3540 		if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
3541 		    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
3542 			if (pa->ifname[0])
3543 				pfi_kkif_unref(pa->kif);
3544 			PF_RULES_WUNLOCK();
3545 			free(pa, M_PFRULE);
3546 			break;
3547 		}
3548 		TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
3549 		PF_RULES_WUNLOCK();
3550 		break;
3551 	}
3552 
3553 	case DIOCGETADDRS: {
3554 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3555 		struct pf_kpool		*pool;
3556 		struct pf_kpooladdr	*pa;
3557 
3558 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
3559 		pp->nr = 0;
3560 
3561 		PF_RULES_RLOCK();
3562 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3563 		    pp->r_num, 0, 1, 0);
3564 		if (pool == NULL) {
3565 			PF_RULES_RUNLOCK();
3566 			error = EBUSY;
3567 			break;
3568 		}
3569 		TAILQ_FOREACH(pa, &pool->list, entries)
3570 			pp->nr++;
3571 		PF_RULES_RUNLOCK();
3572 		break;
3573 	}
3574 
3575 	case DIOCGETADDR: {
3576 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
3577 		struct pf_kpool		*pool;
3578 		struct pf_kpooladdr	*pa;
3579 		u_int32_t		 nr = 0;
3580 
3581 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
3582 
3583 		PF_RULES_RLOCK();
3584 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
3585 		    pp->r_num, 0, 1, 1);
3586 		if (pool == NULL) {
3587 			PF_RULES_RUNLOCK();
3588 			error = EBUSY;
3589 			break;
3590 		}
3591 		pa = TAILQ_FIRST(&pool->list);
3592 		while ((pa != NULL) && (nr < pp->nr)) {
3593 			pa = TAILQ_NEXT(pa, entries);
3594 			nr++;
3595 		}
3596 		if (pa == NULL) {
3597 			PF_RULES_RUNLOCK();
3598 			error = EBUSY;
3599 			break;
3600 		}
3601 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
3602 		pf_addr_copyout(&pp->addr.addr);
3603 		PF_RULES_RUNLOCK();
3604 		break;
3605 	}
3606 
3607 	case DIOCCHANGEADDR: {
3608 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
3609 		struct pf_kpool		*pool;
3610 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
3611 		struct pf_kruleset	*ruleset;
3612 		struct pfi_kkif		*kif = NULL;
3613 
3614 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
3615 
3616 		if (pca->action < PF_CHANGE_ADD_HEAD ||
3617 		    pca->action > PF_CHANGE_REMOVE) {
3618 			error = EINVAL;
3619 			break;
3620 		}
3621 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
3622 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
3623 		    pca->addr.addr.type != PF_ADDR_TABLE) {
3624 			error = EINVAL;
3625 			break;
3626 		}
3627 		if (pca->addr.addr.p.dyn != NULL) {
3628 			error = EINVAL;
3629 			break;
3630 		}
3631 
3632 		if (pca->action != PF_CHANGE_REMOVE) {
3633 #ifndef INET
3634 			if (pca->af == AF_INET) {
3635 				error = EAFNOSUPPORT;
3636 				break;
3637 			}
3638 #endif /* INET */
3639 #ifndef INET6
3640 			if (pca->af == AF_INET6) {
3641 				error = EAFNOSUPPORT;
3642 				break;
3643 			}
3644 #endif /* INET6 */
3645 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
3646 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
3647 			if (newpa->ifname[0])
3648 				kif = pf_kkif_create(M_WAITOK);
3649 			newpa->kif = NULL;
3650 		}
3651 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
3652 		PF_RULES_WLOCK();
3653 		ruleset = pf_find_kruleset(pca->anchor);
3654 		if (ruleset == NULL)
3655 			ERROUT(EBUSY);
3656 
3657 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
3658 		    pca->r_num, pca->r_last, 1, 1);
3659 		if (pool == NULL)
3660 			ERROUT(EBUSY);
3661 
3662 		if (pca->action != PF_CHANGE_REMOVE) {
3663 			if (newpa->ifname[0]) {
3664 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
3665 				pfi_kkif_ref(newpa->kif);
3666 				kif = NULL;
3667 			}
3668 
3669 			switch (newpa->addr.type) {
3670 			case PF_ADDR_DYNIFTL:
3671 				error = pfi_dynaddr_setup(&newpa->addr,
3672 				    pca->af);
3673 				break;
3674 			case PF_ADDR_TABLE:
3675 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
3676 				    newpa->addr.v.tblname);
3677 				if (newpa->addr.p.tbl == NULL)
3678 					error = ENOMEM;
3679 				break;
3680 			}
3681 			if (error)
3682 				goto DIOCCHANGEADDR_error;
3683 		}
3684 
3685 		switch (pca->action) {
3686 		case PF_CHANGE_ADD_HEAD:
3687 			oldpa = TAILQ_FIRST(&pool->list);
3688 			break;
3689 		case PF_CHANGE_ADD_TAIL:
3690 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
3691 			break;
3692 		default:
3693 			oldpa = TAILQ_FIRST(&pool->list);
3694 			for (int i = 0; oldpa && i < pca->nr; i++)
3695 				oldpa = TAILQ_NEXT(oldpa, entries);
3696 
3697 			if (oldpa == NULL)
3698 				ERROUT(EINVAL);
3699 		}
3700 
3701 		if (pca->action == PF_CHANGE_REMOVE) {
3702 			TAILQ_REMOVE(&pool->list, oldpa, entries);
3703 			switch (oldpa->addr.type) {
3704 			case PF_ADDR_DYNIFTL:
3705 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
3706 				break;
3707 			case PF_ADDR_TABLE:
3708 				pfr_detach_table(oldpa->addr.p.tbl);
3709 				break;
3710 			}
3711 			if (oldpa->kif)
3712 				pfi_kkif_unref(oldpa->kif);
3713 			free(oldpa, M_PFRULE);
3714 		} else {
3715 			if (oldpa == NULL)
3716 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
3717 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
3718 			    pca->action == PF_CHANGE_ADD_BEFORE)
3719 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
3720 			else
3721 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
3722 				    newpa, entries);
3723 		}
3724 
3725 		pool->cur = TAILQ_FIRST(&pool->list);
3726 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
3727 		PF_RULES_WUNLOCK();
3728 		break;
3729 
3730 #undef ERROUT
3731 DIOCCHANGEADDR_error:
3732 		if (newpa != NULL) {
3733 			if (newpa->kif)
3734 				pfi_kkif_unref(newpa->kif);
3735 			free(newpa, M_PFRULE);
3736 		}
3737 		PF_RULES_WUNLOCK();
3738 		pf_kkif_free(kif);
3739 		break;
3740 	}
3741 
3742 	case DIOCGETRULESETS: {
3743 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3744 		struct pf_kruleset	*ruleset;
3745 		struct pf_kanchor	*anchor;
3746 
3747 		pr->path[sizeof(pr->path) - 1] = 0;
3748 
3749 		PF_RULES_RLOCK();
3750 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3751 			PF_RULES_RUNLOCK();
3752 			error = ENOENT;
3753 			break;
3754 		}
3755 		pr->nr = 0;
3756 		if (ruleset->anchor == NULL) {
3757 			/* XXX kludge for pf_main_ruleset */
3758 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3759 				if (anchor->parent == NULL)
3760 					pr->nr++;
3761 		} else {
3762 			RB_FOREACH(anchor, pf_kanchor_node,
3763 			    &ruleset->anchor->children)
3764 				pr->nr++;
3765 		}
3766 		PF_RULES_RUNLOCK();
3767 		break;
3768 	}
3769 
3770 	case DIOCGETRULESET: {
3771 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
3772 		struct pf_kruleset	*ruleset;
3773 		struct pf_kanchor	*anchor;
3774 		u_int32_t		 nr = 0;
3775 
3776 		pr->path[sizeof(pr->path) - 1] = 0;
3777 
3778 		PF_RULES_RLOCK();
3779 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
3780 			PF_RULES_RUNLOCK();
3781 			error = ENOENT;
3782 			break;
3783 		}
3784 		pr->name[0] = 0;
3785 		if (ruleset->anchor == NULL) {
3786 			/* XXX kludge for pf_main_ruleset */
3787 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
3788 				if (anchor->parent == NULL && nr++ == pr->nr) {
3789 					strlcpy(pr->name, anchor->name,
3790 					    sizeof(pr->name));
3791 					break;
3792 				}
3793 		} else {
3794 			RB_FOREACH(anchor, pf_kanchor_node,
3795 			    &ruleset->anchor->children)
3796 				if (nr++ == pr->nr) {
3797 					strlcpy(pr->name, anchor->name,
3798 					    sizeof(pr->name));
3799 					break;
3800 				}
3801 		}
3802 		if (!pr->name[0])
3803 			error = EBUSY;
3804 		PF_RULES_RUNLOCK();
3805 		break;
3806 	}
3807 
3808 	case DIOCRCLRTABLES: {
3809 		struct pfioc_table *io = (struct pfioc_table *)addr;
3810 
3811 		if (io->pfrio_esize != 0) {
3812 			error = ENODEV;
3813 			break;
3814 		}
3815 		PF_RULES_WLOCK();
3816 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
3817 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
3818 		PF_RULES_WUNLOCK();
3819 		break;
3820 	}
3821 
3822 	case DIOCRADDTABLES: {
3823 		struct pfioc_table *io = (struct pfioc_table *)addr;
3824 		struct pfr_table *pfrts;
3825 		size_t totlen;
3826 
3827 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3828 			error = ENODEV;
3829 			break;
3830 		}
3831 
3832 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3833 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3834 			error = ENOMEM;
3835 			break;
3836 		}
3837 
3838 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3839 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3840 		    M_TEMP, M_WAITOK);
3841 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3842 		if (error) {
3843 			free(pfrts, M_TEMP);
3844 			break;
3845 		}
3846 		PF_RULES_WLOCK();
3847 		error = pfr_add_tables(pfrts, io->pfrio_size,
3848 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3849 		PF_RULES_WUNLOCK();
3850 		free(pfrts, M_TEMP);
3851 		break;
3852 	}
3853 
3854 	case DIOCRDELTABLES: {
3855 		struct pfioc_table *io = (struct pfioc_table *)addr;
3856 		struct pfr_table *pfrts;
3857 		size_t totlen;
3858 
3859 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3860 			error = ENODEV;
3861 			break;
3862 		}
3863 
3864 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3865 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3866 			error = ENOMEM;
3867 			break;
3868 		}
3869 
3870 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3871 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3872 		    M_TEMP, M_WAITOK);
3873 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3874 		if (error) {
3875 			free(pfrts, M_TEMP);
3876 			break;
3877 		}
3878 		PF_RULES_WLOCK();
3879 		error = pfr_del_tables(pfrts, io->pfrio_size,
3880 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3881 		PF_RULES_WUNLOCK();
3882 		free(pfrts, M_TEMP);
3883 		break;
3884 	}
3885 
3886 	case DIOCRGETTABLES: {
3887 		struct pfioc_table *io = (struct pfioc_table *)addr;
3888 		struct pfr_table *pfrts;
3889 		size_t totlen;
3890 		int n;
3891 
3892 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3893 			error = ENODEV;
3894 			break;
3895 		}
3896 		PF_RULES_RLOCK();
3897 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3898 		if (n < 0) {
3899 			PF_RULES_RUNLOCK();
3900 			error = EINVAL;
3901 			break;
3902 		}
3903 		io->pfrio_size = min(io->pfrio_size, n);
3904 
3905 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3906 
3907 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3908 		    M_TEMP, M_NOWAIT | M_ZERO);
3909 		if (pfrts == NULL) {
3910 			error = ENOMEM;
3911 			PF_RULES_RUNLOCK();
3912 			break;
3913 		}
3914 		error = pfr_get_tables(&io->pfrio_table, pfrts,
3915 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3916 		PF_RULES_RUNLOCK();
3917 		if (error == 0)
3918 			error = copyout(pfrts, io->pfrio_buffer, totlen);
3919 		free(pfrts, M_TEMP);
3920 		break;
3921 	}
3922 
3923 	case DIOCRGETTSTATS: {
3924 		struct pfioc_table *io = (struct pfioc_table *)addr;
3925 		struct pfr_tstats *pfrtstats;
3926 		size_t totlen;
3927 		int n;
3928 
3929 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
3930 			error = ENODEV;
3931 			break;
3932 		}
3933 		PF_TABLE_STATS_LOCK();
3934 		PF_RULES_RLOCK();
3935 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
3936 		if (n < 0) {
3937 			PF_RULES_RUNLOCK();
3938 			PF_TABLE_STATS_UNLOCK();
3939 			error = EINVAL;
3940 			break;
3941 		}
3942 		io->pfrio_size = min(io->pfrio_size, n);
3943 
3944 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
3945 		pfrtstats = mallocarray(io->pfrio_size,
3946 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
3947 		if (pfrtstats == NULL) {
3948 			error = ENOMEM;
3949 			PF_RULES_RUNLOCK();
3950 			PF_TABLE_STATS_UNLOCK();
3951 			break;
3952 		}
3953 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
3954 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3955 		PF_RULES_RUNLOCK();
3956 		PF_TABLE_STATS_UNLOCK();
3957 		if (error == 0)
3958 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
3959 		free(pfrtstats, M_TEMP);
3960 		break;
3961 	}
3962 
3963 	case DIOCRCLRTSTATS: {
3964 		struct pfioc_table *io = (struct pfioc_table *)addr;
3965 		struct pfr_table *pfrts;
3966 		size_t totlen;
3967 
3968 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
3969 			error = ENODEV;
3970 			break;
3971 		}
3972 
3973 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
3974 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
3975 			/* We used to count tables and use the minimum required
3976 			 * size, so we didn't fail on overly large requests.
3977 			 * Keep doing so. */
3978 			io->pfrio_size = pf_ioctl_maxcount;
3979 			break;
3980 		}
3981 
3982 		totlen = io->pfrio_size * sizeof(struct pfr_table);
3983 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
3984 		    M_TEMP, M_WAITOK);
3985 		error = copyin(io->pfrio_buffer, pfrts, totlen);
3986 		if (error) {
3987 			free(pfrts, M_TEMP);
3988 			break;
3989 		}
3990 
3991 		PF_TABLE_STATS_LOCK();
3992 		PF_RULES_RLOCK();
3993 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
3994 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
3995 		PF_RULES_RUNLOCK();
3996 		PF_TABLE_STATS_UNLOCK();
3997 		free(pfrts, M_TEMP);
3998 		break;
3999 	}
4000 
4001 	case DIOCRSETTFLAGS: {
4002 		struct pfioc_table *io = (struct pfioc_table *)addr;
4003 		struct pfr_table *pfrts;
4004 		size_t totlen;
4005 		int n;
4006 
4007 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4008 			error = ENODEV;
4009 			break;
4010 		}
4011 
4012 		PF_RULES_RLOCK();
4013 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4014 		if (n < 0) {
4015 			PF_RULES_RUNLOCK();
4016 			error = EINVAL;
4017 			break;
4018 		}
4019 
4020 		io->pfrio_size = min(io->pfrio_size, n);
4021 		PF_RULES_RUNLOCK();
4022 
4023 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4024 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4025 		    M_TEMP, M_WAITOK);
4026 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4027 		if (error) {
4028 			free(pfrts, M_TEMP);
4029 			break;
4030 		}
4031 		PF_RULES_WLOCK();
4032 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4033 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4034 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4035 		PF_RULES_WUNLOCK();
4036 		free(pfrts, M_TEMP);
4037 		break;
4038 	}
4039 
4040 	case DIOCRCLRADDRS: {
4041 		struct pfioc_table *io = (struct pfioc_table *)addr;
4042 
4043 		if (io->pfrio_esize != 0) {
4044 			error = ENODEV;
4045 			break;
4046 		}
4047 		PF_RULES_WLOCK();
4048 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4049 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4050 		PF_RULES_WUNLOCK();
4051 		break;
4052 	}
4053 
4054 	case DIOCRADDADDRS: {
4055 		struct pfioc_table *io = (struct pfioc_table *)addr;
4056 		struct pfr_addr *pfras;
4057 		size_t totlen;
4058 
4059 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4060 			error = ENODEV;
4061 			break;
4062 		}
4063 		if (io->pfrio_size < 0 ||
4064 		    io->pfrio_size > pf_ioctl_maxcount ||
4065 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4066 			error = EINVAL;
4067 			break;
4068 		}
4069 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4070 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4071 		    M_TEMP, M_WAITOK);
4072 		error = copyin(io->pfrio_buffer, pfras, totlen);
4073 		if (error) {
4074 			free(pfras, M_TEMP);
4075 			break;
4076 		}
4077 		PF_RULES_WLOCK();
4078 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4079 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4080 		    PFR_FLAG_USERIOCTL);
4081 		PF_RULES_WUNLOCK();
4082 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4083 			error = copyout(pfras, io->pfrio_buffer, totlen);
4084 		free(pfras, M_TEMP);
4085 		break;
4086 	}
4087 
4088 	case DIOCRDELADDRS: {
4089 		struct pfioc_table *io = (struct pfioc_table *)addr;
4090 		struct pfr_addr *pfras;
4091 		size_t totlen;
4092 
4093 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4094 			error = ENODEV;
4095 			break;
4096 		}
4097 		if (io->pfrio_size < 0 ||
4098 		    io->pfrio_size > pf_ioctl_maxcount ||
4099 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4100 			error = EINVAL;
4101 			break;
4102 		}
4103 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4104 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4105 		    M_TEMP, M_WAITOK);
4106 		error = copyin(io->pfrio_buffer, pfras, totlen);
4107 		if (error) {
4108 			free(pfras, M_TEMP);
4109 			break;
4110 		}
4111 		PF_RULES_WLOCK();
4112 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4113 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4114 		    PFR_FLAG_USERIOCTL);
4115 		PF_RULES_WUNLOCK();
4116 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4117 			error = copyout(pfras, io->pfrio_buffer, totlen);
4118 		free(pfras, M_TEMP);
4119 		break;
4120 	}
4121 
4122 	case DIOCRSETADDRS: {
4123 		struct pfioc_table *io = (struct pfioc_table *)addr;
4124 		struct pfr_addr *pfras;
4125 		size_t totlen, count;
4126 
4127 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4128 			error = ENODEV;
4129 			break;
4130 		}
4131 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4132 			error = EINVAL;
4133 			break;
4134 		}
4135 		count = max(io->pfrio_size, io->pfrio_size2);
4136 		if (count > pf_ioctl_maxcount ||
4137 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4138 			error = EINVAL;
4139 			break;
4140 		}
4141 		totlen = count * sizeof(struct pfr_addr);
4142 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4143 		    M_WAITOK);
4144 		error = copyin(io->pfrio_buffer, pfras, totlen);
4145 		if (error) {
4146 			free(pfras, M_TEMP);
4147 			break;
4148 		}
4149 		PF_RULES_WLOCK();
4150 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4151 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4152 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4153 		    PFR_FLAG_USERIOCTL, 0);
4154 		PF_RULES_WUNLOCK();
4155 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4156 			error = copyout(pfras, io->pfrio_buffer, totlen);
4157 		free(pfras, M_TEMP);
4158 		break;
4159 	}
4160 
4161 	case DIOCRGETADDRS: {
4162 		struct pfioc_table *io = (struct pfioc_table *)addr;
4163 		struct pfr_addr *pfras;
4164 		size_t totlen;
4165 
4166 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4167 			error = ENODEV;
4168 			break;
4169 		}
4170 		if (io->pfrio_size < 0 ||
4171 		    io->pfrio_size > pf_ioctl_maxcount ||
4172 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4173 			error = EINVAL;
4174 			break;
4175 		}
4176 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4177 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4178 		    M_TEMP, M_WAITOK | M_ZERO);
4179 		PF_RULES_RLOCK();
4180 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4181 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4182 		PF_RULES_RUNLOCK();
4183 		if (error == 0)
4184 			error = copyout(pfras, io->pfrio_buffer, totlen);
4185 		free(pfras, M_TEMP);
4186 		break;
4187 	}
4188 
4189 	case DIOCRGETASTATS: {
4190 		struct pfioc_table *io = (struct pfioc_table *)addr;
4191 		struct pfr_astats *pfrastats;
4192 		size_t totlen;
4193 
4194 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4195 			error = ENODEV;
4196 			break;
4197 		}
4198 		if (io->pfrio_size < 0 ||
4199 		    io->pfrio_size > pf_ioctl_maxcount ||
4200 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4201 			error = EINVAL;
4202 			break;
4203 		}
4204 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4205 		pfrastats = mallocarray(io->pfrio_size,
4206 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4207 		PF_RULES_RLOCK();
4208 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4209 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4210 		PF_RULES_RUNLOCK();
4211 		if (error == 0)
4212 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4213 		free(pfrastats, M_TEMP);
4214 		break;
4215 	}
4216 
4217 	case DIOCRCLRASTATS: {
4218 		struct pfioc_table *io = (struct pfioc_table *)addr;
4219 		struct pfr_addr *pfras;
4220 		size_t totlen;
4221 
4222 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4223 			error = ENODEV;
4224 			break;
4225 		}
4226 		if (io->pfrio_size < 0 ||
4227 		    io->pfrio_size > pf_ioctl_maxcount ||
4228 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4229 			error = EINVAL;
4230 			break;
4231 		}
4232 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4233 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4234 		    M_TEMP, M_WAITOK);
4235 		error = copyin(io->pfrio_buffer, pfras, totlen);
4236 		if (error) {
4237 			free(pfras, M_TEMP);
4238 			break;
4239 		}
4240 		PF_RULES_WLOCK();
4241 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4242 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4243 		    PFR_FLAG_USERIOCTL);
4244 		PF_RULES_WUNLOCK();
4245 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4246 			error = copyout(pfras, io->pfrio_buffer, totlen);
4247 		free(pfras, M_TEMP);
4248 		break;
4249 	}
4250 
4251 	case DIOCRTSTADDRS: {
4252 		struct pfioc_table *io = (struct pfioc_table *)addr;
4253 		struct pfr_addr *pfras;
4254 		size_t totlen;
4255 
4256 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4257 			error = ENODEV;
4258 			break;
4259 		}
4260 		if (io->pfrio_size < 0 ||
4261 		    io->pfrio_size > pf_ioctl_maxcount ||
4262 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4263 			error = EINVAL;
4264 			break;
4265 		}
4266 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4267 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4268 		    M_TEMP, M_WAITOK);
4269 		error = copyin(io->pfrio_buffer, pfras, totlen);
4270 		if (error) {
4271 			free(pfras, M_TEMP);
4272 			break;
4273 		}
4274 		PF_RULES_RLOCK();
4275 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
4276 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
4277 		    PFR_FLAG_USERIOCTL);
4278 		PF_RULES_RUNLOCK();
4279 		if (error == 0)
4280 			error = copyout(pfras, io->pfrio_buffer, totlen);
4281 		free(pfras, M_TEMP);
4282 		break;
4283 	}
4284 
4285 	case DIOCRINADEFINE: {
4286 		struct pfioc_table *io = (struct pfioc_table *)addr;
4287 		struct pfr_addr *pfras;
4288 		size_t totlen;
4289 
4290 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4291 			error = ENODEV;
4292 			break;
4293 		}
4294 		if (io->pfrio_size < 0 ||
4295 		    io->pfrio_size > pf_ioctl_maxcount ||
4296 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4297 			error = EINVAL;
4298 			break;
4299 		}
4300 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4301 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4302 		    M_TEMP, M_WAITOK);
4303 		error = copyin(io->pfrio_buffer, pfras, totlen);
4304 		if (error) {
4305 			free(pfras, M_TEMP);
4306 			break;
4307 		}
4308 		PF_RULES_WLOCK();
4309 		error = pfr_ina_define(&io->pfrio_table, pfras,
4310 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
4311 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4312 		PF_RULES_WUNLOCK();
4313 		free(pfras, M_TEMP);
4314 		break;
4315 	}
4316 
4317 	case DIOCOSFPADD: {
4318 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4319 		PF_RULES_WLOCK();
4320 		error = pf_osfp_add(io);
4321 		PF_RULES_WUNLOCK();
4322 		break;
4323 	}
4324 
4325 	case DIOCOSFPGET: {
4326 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
4327 		PF_RULES_RLOCK();
4328 		error = pf_osfp_get(io);
4329 		PF_RULES_RUNLOCK();
4330 		break;
4331 	}
4332 
4333 	case DIOCXBEGIN: {
4334 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4335 		struct pfioc_trans_e	*ioes, *ioe;
4336 		size_t			 totlen;
4337 		int			 i;
4338 
4339 		if (io->esize != sizeof(*ioe)) {
4340 			error = ENODEV;
4341 			break;
4342 		}
4343 		if (io->size < 0 ||
4344 		    io->size > pf_ioctl_maxcount ||
4345 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4346 			error = EINVAL;
4347 			break;
4348 		}
4349 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4350 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4351 		    M_TEMP, M_WAITOK);
4352 		error = copyin(io->array, ioes, totlen);
4353 		if (error) {
4354 			free(ioes, M_TEMP);
4355 			break;
4356 		}
4357 		PF_RULES_WLOCK();
4358 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4359 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
4360 			switch (ioe->rs_num) {
4361 #ifdef ALTQ
4362 			case PF_RULESET_ALTQ:
4363 				if (ioe->anchor[0]) {
4364 					PF_RULES_WUNLOCK();
4365 					free(ioes, M_TEMP);
4366 					error = EINVAL;
4367 					goto fail;
4368 				}
4369 				if ((error = pf_begin_altq(&ioe->ticket))) {
4370 					PF_RULES_WUNLOCK();
4371 					free(ioes, M_TEMP);
4372 					goto fail;
4373 				}
4374 				break;
4375 #endif /* ALTQ */
4376 			case PF_RULESET_TABLE:
4377 			    {
4378 				struct pfr_table table;
4379 
4380 				bzero(&table, sizeof(table));
4381 				strlcpy(table.pfrt_anchor, ioe->anchor,
4382 				    sizeof(table.pfrt_anchor));
4383 				if ((error = pfr_ina_begin(&table,
4384 				    &ioe->ticket, NULL, 0))) {
4385 					PF_RULES_WUNLOCK();
4386 					free(ioes, M_TEMP);
4387 					goto fail;
4388 				}
4389 				break;
4390 			    }
4391 			default:
4392 				if ((error = pf_begin_rules(&ioe->ticket,
4393 				    ioe->rs_num, ioe->anchor))) {
4394 					PF_RULES_WUNLOCK();
4395 					free(ioes, M_TEMP);
4396 					goto fail;
4397 				}
4398 				break;
4399 			}
4400 		}
4401 		PF_RULES_WUNLOCK();
4402 		error = copyout(ioes, io->array, totlen);
4403 		free(ioes, M_TEMP);
4404 		break;
4405 	}
4406 
4407 	case DIOCXROLLBACK: {
4408 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4409 		struct pfioc_trans_e	*ioe, *ioes;
4410 		size_t			 totlen;
4411 		int			 i;
4412 
4413 		if (io->esize != sizeof(*ioe)) {
4414 			error = ENODEV;
4415 			break;
4416 		}
4417 		if (io->size < 0 ||
4418 		    io->size > pf_ioctl_maxcount ||
4419 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4420 			error = EINVAL;
4421 			break;
4422 		}
4423 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4424 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4425 		    M_TEMP, M_WAITOK);
4426 		error = copyin(io->array, ioes, totlen);
4427 		if (error) {
4428 			free(ioes, M_TEMP);
4429 			break;
4430 		}
4431 		PF_RULES_WLOCK();
4432 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4433 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
4434 			switch (ioe->rs_num) {
4435 #ifdef ALTQ
4436 			case PF_RULESET_ALTQ:
4437 				if (ioe->anchor[0]) {
4438 					PF_RULES_WUNLOCK();
4439 					free(ioes, M_TEMP);
4440 					error = EINVAL;
4441 					goto fail;
4442 				}
4443 				if ((error = pf_rollback_altq(ioe->ticket))) {
4444 					PF_RULES_WUNLOCK();
4445 					free(ioes, M_TEMP);
4446 					goto fail; /* really bad */
4447 				}
4448 				break;
4449 #endif /* ALTQ */
4450 			case PF_RULESET_TABLE:
4451 			    {
4452 				struct pfr_table table;
4453 
4454 				bzero(&table, sizeof(table));
4455 				strlcpy(table.pfrt_anchor, ioe->anchor,
4456 				    sizeof(table.pfrt_anchor));
4457 				if ((error = pfr_ina_rollback(&table,
4458 				    ioe->ticket, NULL, 0))) {
4459 					PF_RULES_WUNLOCK();
4460 					free(ioes, M_TEMP);
4461 					goto fail; /* really bad */
4462 				}
4463 				break;
4464 			    }
4465 			default:
4466 				if ((error = pf_rollback_rules(ioe->ticket,
4467 				    ioe->rs_num, ioe->anchor))) {
4468 					PF_RULES_WUNLOCK();
4469 					free(ioes, M_TEMP);
4470 					goto fail; /* really bad */
4471 				}
4472 				break;
4473 			}
4474 		}
4475 		PF_RULES_WUNLOCK();
4476 		free(ioes, M_TEMP);
4477 		break;
4478 	}
4479 
4480 	case DIOCXCOMMIT: {
4481 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
4482 		struct pfioc_trans_e	*ioe, *ioes;
4483 		struct pf_kruleset	*rs;
4484 		size_t			 totlen;
4485 		int			 i;
4486 
4487 		if (io->esize != sizeof(*ioe)) {
4488 			error = ENODEV;
4489 			break;
4490 		}
4491 
4492 		if (io->size < 0 ||
4493 		    io->size > pf_ioctl_maxcount ||
4494 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
4495 			error = EINVAL;
4496 			break;
4497 		}
4498 
4499 		totlen = sizeof(struct pfioc_trans_e) * io->size;
4500 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
4501 		    M_TEMP, M_WAITOK);
4502 		error = copyin(io->array, ioes, totlen);
4503 		if (error) {
4504 			free(ioes, M_TEMP);
4505 			break;
4506 		}
4507 		PF_RULES_WLOCK();
4508 		/* First makes sure everything will succeed. */
4509 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4510 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
4511 			switch (ioe->rs_num) {
4512 #ifdef ALTQ
4513 			case PF_RULESET_ALTQ:
4514 				if (ioe->anchor[0]) {
4515 					PF_RULES_WUNLOCK();
4516 					free(ioes, M_TEMP);
4517 					error = EINVAL;
4518 					goto fail;
4519 				}
4520 				if (!V_altqs_inactive_open || ioe->ticket !=
4521 				    V_ticket_altqs_inactive) {
4522 					PF_RULES_WUNLOCK();
4523 					free(ioes, M_TEMP);
4524 					error = EBUSY;
4525 					goto fail;
4526 				}
4527 				break;
4528 #endif /* ALTQ */
4529 			case PF_RULESET_TABLE:
4530 				rs = pf_find_kruleset(ioe->anchor);
4531 				if (rs == NULL || !rs->topen || ioe->ticket !=
4532 				    rs->tticket) {
4533 					PF_RULES_WUNLOCK();
4534 					free(ioes, M_TEMP);
4535 					error = EBUSY;
4536 					goto fail;
4537 				}
4538 				break;
4539 			default:
4540 				if (ioe->rs_num < 0 || ioe->rs_num >=
4541 				    PF_RULESET_MAX) {
4542 					PF_RULES_WUNLOCK();
4543 					free(ioes, M_TEMP);
4544 					error = EINVAL;
4545 					goto fail;
4546 				}
4547 				rs = pf_find_kruleset(ioe->anchor);
4548 				if (rs == NULL ||
4549 				    !rs->rules[ioe->rs_num].inactive.open ||
4550 				    rs->rules[ioe->rs_num].inactive.ticket !=
4551 				    ioe->ticket) {
4552 					PF_RULES_WUNLOCK();
4553 					free(ioes, M_TEMP);
4554 					error = EBUSY;
4555 					goto fail;
4556 				}
4557 				break;
4558 			}
4559 		}
4560 		/* Now do the commit - no errors should happen here. */
4561 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
4562 			switch (ioe->rs_num) {
4563 #ifdef ALTQ
4564 			case PF_RULESET_ALTQ:
4565 				if ((error = pf_commit_altq(ioe->ticket))) {
4566 					PF_RULES_WUNLOCK();
4567 					free(ioes, M_TEMP);
4568 					goto fail; /* really bad */
4569 				}
4570 				break;
4571 #endif /* ALTQ */
4572 			case PF_RULESET_TABLE:
4573 			    {
4574 				struct pfr_table table;
4575 
4576 				bzero(&table, sizeof(table));
4577 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
4578 				    sizeof(table.pfrt_anchor));
4579 				if ((error = pfr_ina_commit(&table,
4580 				    ioe->ticket, NULL, NULL, 0))) {
4581 					PF_RULES_WUNLOCK();
4582 					free(ioes, M_TEMP);
4583 					goto fail; /* really bad */
4584 				}
4585 				break;
4586 			    }
4587 			default:
4588 				if ((error = pf_commit_rules(ioe->ticket,
4589 				    ioe->rs_num, ioe->anchor))) {
4590 					PF_RULES_WUNLOCK();
4591 					free(ioes, M_TEMP);
4592 					goto fail; /* really bad */
4593 				}
4594 				break;
4595 			}
4596 		}
4597 		PF_RULES_WUNLOCK();
4598 		free(ioes, M_TEMP);
4599 		break;
4600 	}
4601 
4602 	case DIOCGETSRCNODES: {
4603 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
4604 		struct pf_srchash	*sh;
4605 		struct pf_ksrc_node	*n;
4606 		struct pf_src_node	*p, *pstore;
4607 		uint32_t		 i, nr = 0;
4608 
4609 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4610 				i++, sh++) {
4611 			PF_HASHROW_LOCK(sh);
4612 			LIST_FOREACH(n, &sh->nodes, entry)
4613 				nr++;
4614 			PF_HASHROW_UNLOCK(sh);
4615 		}
4616 
4617 		psn->psn_len = min(psn->psn_len,
4618 		    sizeof(struct pf_src_node) * nr);
4619 
4620 		if (psn->psn_len == 0) {
4621 			psn->psn_len = sizeof(struct pf_src_node) * nr;
4622 			break;
4623 		}
4624 
4625 		nr = 0;
4626 
4627 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
4628 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
4629 		    i++, sh++) {
4630 		    PF_HASHROW_LOCK(sh);
4631 		    LIST_FOREACH(n, &sh->nodes, entry) {
4632 
4633 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
4634 				break;
4635 
4636 			pf_src_node_copy(n, p);
4637 
4638 			p++;
4639 			nr++;
4640 		    }
4641 		    PF_HASHROW_UNLOCK(sh);
4642 		}
4643 		error = copyout(pstore, psn->psn_src_nodes,
4644 		    sizeof(struct pf_src_node) * nr);
4645 		if (error) {
4646 			free(pstore, M_TEMP);
4647 			break;
4648 		}
4649 		psn->psn_len = sizeof(struct pf_src_node) * nr;
4650 		free(pstore, M_TEMP);
4651 		break;
4652 	}
4653 
4654 	case DIOCCLRSRCNODES: {
4655 		pf_clear_srcnodes(NULL);
4656 		pf_purge_expired_src_nodes();
4657 		break;
4658 	}
4659 
4660 	case DIOCKILLSRCNODES:
4661 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
4662 		break;
4663 
4664 	case DIOCKEEPCOUNTERS:
4665 		error = pf_keepcounters((struct pfioc_nv *)addr);
4666 		break;
4667 
4668 	case DIOCGETSYNCOOKIES:
4669 		error = pf_get_syncookies((struct pfioc_nv *)addr);
4670 		break;
4671 
4672 	case DIOCSETSYNCOOKIES:
4673 		error = pf_set_syncookies((struct pfioc_nv *)addr);
4674 		break;
4675 
4676 	case DIOCSETHOSTID: {
4677 		u_int32_t	*hostid = (u_int32_t *)addr;
4678 
4679 		PF_RULES_WLOCK();
4680 		if (*hostid == 0)
4681 			V_pf_status.hostid = arc4random();
4682 		else
4683 			V_pf_status.hostid = *hostid;
4684 		PF_RULES_WUNLOCK();
4685 		break;
4686 	}
4687 
4688 	case DIOCOSFPFLUSH:
4689 		PF_RULES_WLOCK();
4690 		pf_osfp_flush();
4691 		PF_RULES_WUNLOCK();
4692 		break;
4693 
4694 	case DIOCIGETIFACES: {
4695 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4696 		struct pfi_kif *ifstore;
4697 		size_t bufsiz;
4698 
4699 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
4700 			error = ENODEV;
4701 			break;
4702 		}
4703 
4704 		if (io->pfiio_size < 0 ||
4705 		    io->pfiio_size > pf_ioctl_maxcount ||
4706 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
4707 			error = EINVAL;
4708 			break;
4709 		}
4710 
4711 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
4712 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
4713 		    M_TEMP, M_WAITOK | M_ZERO);
4714 
4715 		PF_RULES_RLOCK();
4716 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
4717 		PF_RULES_RUNLOCK();
4718 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
4719 		free(ifstore, M_TEMP);
4720 		break;
4721 	}
4722 
4723 	case DIOCSETIFFLAG: {
4724 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4725 
4726 		PF_RULES_WLOCK();
4727 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
4728 		PF_RULES_WUNLOCK();
4729 		break;
4730 	}
4731 
4732 	case DIOCCLRIFFLAG: {
4733 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
4734 
4735 		PF_RULES_WLOCK();
4736 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
4737 		PF_RULES_WUNLOCK();
4738 		break;
4739 	}
4740 
4741 	default:
4742 		error = ENODEV;
4743 		break;
4744 	}
4745 fail:
4746 	if (sx_xlocked(&pf_ioctl_lock))
4747 		sx_xunlock(&pf_ioctl_lock);
4748 	CURVNET_RESTORE();
4749 
4750 #undef ERROUT_IOCTL
4751 
4752 	return (error);
4753 }
4754 
4755 void
4756 pfsync_state_export(struct pfsync_state *sp, struct pf_kstate *st)
4757 {
4758 	bzero(sp, sizeof(struct pfsync_state));
4759 
4760 	/* copy from state key */
4761 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4762 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4763 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4764 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4765 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4766 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4767 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4768 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4769 	sp->proto = st->key[PF_SK_WIRE]->proto;
4770 	sp->af = st->key[PF_SK_WIRE]->af;
4771 
4772 	/* copy from state */
4773 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4774 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4775 	sp->creation = htonl(time_uptime - st->creation);
4776 	sp->expire = pf_state_expires(st);
4777 	if (sp->expire <= time_uptime)
4778 		sp->expire = htonl(0);
4779 	else
4780 		sp->expire = htonl(sp->expire - time_uptime);
4781 
4782 	sp->direction = st->direction;
4783 	sp->log = st->log;
4784 	sp->timeout = st->timeout;
4785 	sp->state_flags = st->state_flags;
4786 	if (st->src_node)
4787 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4788 	if (st->nat_src_node)
4789 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4790 
4791 	sp->id = st->id;
4792 	sp->creatorid = st->creatorid;
4793 	pf_state_peer_hton(&st->src, &sp->src);
4794 	pf_state_peer_hton(&st->dst, &sp->dst);
4795 
4796 	if (st->rule.ptr == NULL)
4797 		sp->rule = htonl(-1);
4798 	else
4799 		sp->rule = htonl(st->rule.ptr->nr);
4800 	if (st->anchor.ptr == NULL)
4801 		sp->anchor = htonl(-1);
4802 	else
4803 		sp->anchor = htonl(st->anchor.ptr->nr);
4804 	if (st->nat_rule.ptr == NULL)
4805 		sp->nat_rule = htonl(-1);
4806 	else
4807 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4808 
4809 	pf_state_counter_hton(st->packets[0], sp->packets[0]);
4810 	pf_state_counter_hton(st->packets[1], sp->packets[1]);
4811 	pf_state_counter_hton(st->bytes[0], sp->bytes[0]);
4812 	pf_state_counter_hton(st->bytes[1], sp->bytes[1]);
4813 }
4814 
4815 void
4816 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
4817 {
4818 	bzero(sp, sizeof(*sp));
4819 
4820 	sp->version = PF_STATE_VERSION;
4821 
4822 	/* copy from state key */
4823 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
4824 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
4825 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
4826 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
4827 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
4828 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
4829 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
4830 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
4831 	sp->proto = st->key[PF_SK_WIRE]->proto;
4832 	sp->af = st->key[PF_SK_WIRE]->af;
4833 
4834 	/* copy from state */
4835 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
4836 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
4837 	    sizeof(sp->orig_ifname));
4838 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
4839 	sp->creation = htonl(time_uptime - st->creation);
4840 	sp->expire = pf_state_expires(st);
4841 	if (sp->expire <= time_uptime)
4842 		sp->expire = htonl(0);
4843 	else
4844 		sp->expire = htonl(sp->expire - time_uptime);
4845 
4846 	sp->direction = st->direction;
4847 	sp->log = st->log;
4848 	sp->timeout = st->timeout;
4849 	sp->state_flags = st->state_flags;
4850 	if (st->src_node)
4851 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
4852 	if (st->nat_src_node)
4853 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
4854 
4855 	sp->id = st->id;
4856 	sp->creatorid = st->creatorid;
4857 	pf_state_peer_hton(&st->src, &sp->src);
4858 	pf_state_peer_hton(&st->dst, &sp->dst);
4859 
4860 	if (st->rule.ptr == NULL)
4861 		sp->rule = htonl(-1);
4862 	else
4863 		sp->rule = htonl(st->rule.ptr->nr);
4864 	if (st->anchor.ptr == NULL)
4865 		sp->anchor = htonl(-1);
4866 	else
4867 		sp->anchor = htonl(st->anchor.ptr->nr);
4868 	if (st->nat_rule.ptr == NULL)
4869 		sp->nat_rule = htonl(-1);
4870 	else
4871 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
4872 
4873 	sp->packets[0] = st->packets[0];
4874 	sp->packets[1] = st->packets[1];
4875 	sp->bytes[0] = st->bytes[0];
4876 	sp->bytes[1] = st->bytes[1];
4877 }
4878 
4879 static void
4880 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
4881 {
4882 	struct pfr_ktable *kt;
4883 
4884 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
4885 
4886 	kt = aw->p.tbl;
4887 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
4888 		kt = kt->pfrkt_root;
4889 	aw->p.tbl = NULL;
4890 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
4891 		kt->pfrkt_cnt : -1;
4892 }
4893 
4894 static int
4895 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
4896     size_t number, char **names)
4897 {
4898 	nvlist_t        *nvc;
4899 
4900 	nvc = nvlist_create(0);
4901 	if (nvc == NULL)
4902 		return (ENOMEM);
4903 
4904 	for (int i = 0; i < number; i++) {
4905 		nvlist_append_number_array(nvc, "counters",
4906 		    counter_u64_fetch(counters[i]));
4907 		nvlist_append_string_array(nvc, "names",
4908 		    names[i]);
4909 		nvlist_append_number_array(nvc, "ids",
4910 		    i);
4911 	}
4912 	nvlist_add_nvlist(nvl, name, nvc);
4913 	nvlist_destroy(nvc);
4914 
4915 	return (0);
4916 }
4917 
4918 static int
4919 pf_getstatus(struct pfioc_nv *nv)
4920 {
4921 	nvlist_t        *nvl = NULL, *nvc = NULL;
4922 	void            *nvlpacked = NULL;
4923 	int              error;
4924 	struct pf_status s;
4925 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
4926 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
4927 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
4928 	PF_RULES_RLOCK_TRACKER;
4929 
4930 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
4931 
4932 	PF_RULES_RLOCK();
4933 
4934 	nvl = nvlist_create(0);
4935 	if (nvl == NULL)
4936 		ERROUT(ENOMEM);
4937 
4938 	nvlist_add_bool(nvl, "running", V_pf_status.running);
4939 	nvlist_add_number(nvl, "since", V_pf_status.since);
4940 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
4941 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
4942 	nvlist_add_number(nvl, "states", V_pf_status.states);
4943 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
4944 
4945 	/* counters */
4946 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
4947 	    PFRES_MAX, pf_reasons);
4948 	if (error != 0)
4949 		ERROUT(error);
4950 
4951 	/* lcounters */
4952 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
4953 	    KLCNT_MAX, pf_lcounter);
4954 	if (error != 0)
4955 		ERROUT(error);
4956 
4957 	/* fcounters */
4958 	nvc = nvlist_create(0);
4959 	if (nvc == NULL)
4960 		ERROUT(ENOMEM);
4961 
4962 	for (int i = 0; i < FCNT_MAX; i++) {
4963 		nvlist_append_number_array(nvc, "counters",
4964 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
4965 		nvlist_append_string_array(nvc, "names",
4966 		    pf_fcounter[i]);
4967 		nvlist_append_number_array(nvc, "ids",
4968 		    i);
4969 	}
4970 	nvlist_add_nvlist(nvl, "fcounters", nvc);
4971 	nvlist_destroy(nvc);
4972 	nvc = NULL;
4973 
4974 	/* scounters */
4975 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
4976 	    SCNT_MAX, pf_fcounter);
4977 	if (error != 0)
4978 		ERROUT(error);
4979 
4980 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
4981 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
4982 	    PF_MD5_DIGEST_LENGTH);
4983 
4984 	pfi_update_status(V_pf_status.ifname, &s);
4985 
4986 	/* pcounters / bcounters */
4987 	for (int i = 0; i < 2; i++) {
4988 		for (int j = 0; j < 2; j++) {
4989 			for (int k = 0; k < 2; k++) {
4990 				nvlist_append_number_array(nvl, "pcounters",
4991 				    s.pcounters[i][j][k]);
4992 			}
4993 			nvlist_append_number_array(nvl, "bcounters",
4994 			    s.bcounters[i][j]);
4995 		}
4996 	}
4997 
4998 	nvlpacked = nvlist_pack(nvl, &nv->len);
4999 	if (nvlpacked == NULL)
5000 		ERROUT(ENOMEM);
5001 
5002 	if (nv->size == 0)
5003 		ERROUT(0);
5004 	else if (nv->size < nv->len)
5005 		ERROUT(ENOSPC);
5006 
5007 	PF_RULES_RUNLOCK();
5008 	error = copyout(nvlpacked, nv->data, nv->len);
5009 	goto done;
5010 
5011 #undef ERROUT
5012 errout:
5013 	PF_RULES_RUNLOCK();
5014 done:
5015 	free(nvlpacked, M_NVLIST);
5016 	nvlist_destroy(nvc);
5017 	nvlist_destroy(nvl);
5018 
5019 	return (error);
5020 }
5021 
5022 /*
5023  * XXX - Check for version missmatch!!!
5024  */
5025 static void
5026 pf_clear_all_states(void)
5027 {
5028 	struct pf_kstate	*s;
5029 	u_int i;
5030 
5031 	for (i = 0; i <= pf_hashmask; i++) {
5032 		struct pf_idhash *ih = &V_pf_idhash[i];
5033 relock:
5034 		PF_HASHROW_LOCK(ih);
5035 		LIST_FOREACH(s, &ih->states, entry) {
5036 			s->timeout = PFTM_PURGE;
5037 			/* Don't send out individual delete messages. */
5038 			s->state_flags |= PFSTATE_NOSYNC;
5039 			pf_unlink_state(s, PF_ENTER_LOCKED);
5040 			goto relock;
5041 		}
5042 		PF_HASHROW_UNLOCK(ih);
5043 	}
5044 }
5045 
5046 static int
5047 pf_clear_tables(void)
5048 {
5049 	struct pfioc_table io;
5050 	int error;
5051 
5052 	bzero(&io, sizeof(io));
5053 
5054 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5055 	    io.pfrio_flags);
5056 
5057 	return (error);
5058 }
5059 
5060 static void
5061 pf_clear_srcnodes(struct pf_ksrc_node *n)
5062 {
5063 	struct pf_kstate *s;
5064 	int i;
5065 
5066 	for (i = 0; i <= pf_hashmask; i++) {
5067 		struct pf_idhash *ih = &V_pf_idhash[i];
5068 
5069 		PF_HASHROW_LOCK(ih);
5070 		LIST_FOREACH(s, &ih->states, entry) {
5071 			if (n == NULL || n == s->src_node)
5072 				s->src_node = NULL;
5073 			if (n == NULL || n == s->nat_src_node)
5074 				s->nat_src_node = NULL;
5075 		}
5076 		PF_HASHROW_UNLOCK(ih);
5077 	}
5078 
5079 	if (n == NULL) {
5080 		struct pf_srchash *sh;
5081 
5082 		for (i = 0, sh = V_pf_srchash; i <= pf_srchashmask;
5083 		    i++, sh++) {
5084 			PF_HASHROW_LOCK(sh);
5085 			LIST_FOREACH(n, &sh->nodes, entry) {
5086 				n->expire = 1;
5087 				n->states = 0;
5088 			}
5089 			PF_HASHROW_UNLOCK(sh);
5090 		}
5091 	} else {
5092 		/* XXX: hash slot should already be locked here. */
5093 		n->expire = 1;
5094 		n->states = 0;
5095 	}
5096 }
5097 
5098 static void
5099 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5100 {
5101 	struct pf_ksrc_node_list	 kill;
5102 
5103 	LIST_INIT(&kill);
5104 	for (int i = 0; i <= pf_srchashmask; i++) {
5105 		struct pf_srchash *sh = &V_pf_srchash[i];
5106 		struct pf_ksrc_node *sn, *tmp;
5107 
5108 		PF_HASHROW_LOCK(sh);
5109 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5110 			if (PF_MATCHA(psnk->psnk_src.neg,
5111 			      &psnk->psnk_src.addr.v.a.addr,
5112 			      &psnk->psnk_src.addr.v.a.mask,
5113 			      &sn->addr, sn->af) &&
5114 			    PF_MATCHA(psnk->psnk_dst.neg,
5115 			      &psnk->psnk_dst.addr.v.a.addr,
5116 			      &psnk->psnk_dst.addr.v.a.mask,
5117 			      &sn->raddr, sn->af)) {
5118 				pf_unlink_src_node(sn);
5119 				LIST_INSERT_HEAD(&kill, sn, entry);
5120 				sn->expire = 1;
5121 			}
5122 		PF_HASHROW_UNLOCK(sh);
5123 	}
5124 
5125 	for (int i = 0; i <= pf_hashmask; i++) {
5126 		struct pf_idhash *ih = &V_pf_idhash[i];
5127 		struct pf_kstate *s;
5128 
5129 		PF_HASHROW_LOCK(ih);
5130 		LIST_FOREACH(s, &ih->states, entry) {
5131 			if (s->src_node && s->src_node->expire == 1)
5132 				s->src_node = NULL;
5133 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5134 				s->nat_src_node = NULL;
5135 		}
5136 		PF_HASHROW_UNLOCK(ih);
5137 	}
5138 
5139 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5140 }
5141 
5142 static int
5143 pf_keepcounters(struct pfioc_nv *nv)
5144 {
5145 	nvlist_t	*nvl = NULL;
5146 	void		*nvlpacked = NULL;
5147 	int		 error = 0;
5148 
5149 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5150 
5151 	if (nv->len > pf_ioctl_maxcount)
5152 		ERROUT(ENOMEM);
5153 
5154 	nvlpacked = malloc(nv->len, M_TEMP, M_WAITOK);
5155 	if (nvlpacked == NULL)
5156 		ERROUT(ENOMEM);
5157 
5158 	error = copyin(nv->data, nvlpacked, nv->len);
5159 	if (error)
5160 		ERROUT(error);
5161 
5162 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5163 	if (nvl == NULL)
5164 		ERROUT(EBADMSG);
5165 
5166 	if (! nvlist_exists_bool(nvl, "keep_counters"))
5167 		ERROUT(EBADMSG);
5168 
5169 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
5170 
5171 on_error:
5172 	nvlist_destroy(nvl);
5173 	free(nvlpacked, M_TEMP);
5174 	return (error);
5175 }
5176 
5177 static unsigned int
5178 pf_clear_states(const struct pf_kstate_kill *kill)
5179 {
5180 	struct pf_state_key_cmp	 match_key;
5181 	struct pf_kstate	*s;
5182 	struct pfi_kkif	*kif;
5183 	int		 idx;
5184 	unsigned int	 killed = 0, dir;
5185 
5186 	for (unsigned int i = 0; i <= pf_hashmask; i++) {
5187 		struct pf_idhash *ih = &V_pf_idhash[i];
5188 
5189 relock_DIOCCLRSTATES:
5190 		PF_HASHROW_LOCK(ih);
5191 		LIST_FOREACH(s, &ih->states, entry) {
5192 			/* For floating states look at the original kif. */
5193 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
5194 
5195 			if (kill->psk_ifname[0] &&
5196 			    strcmp(kill->psk_ifname,
5197 			    kif->pfik_name))
5198 				continue;
5199 
5200 			if (kill->psk_kill_match) {
5201 				bzero(&match_key, sizeof(match_key));
5202 
5203 				if (s->direction == PF_OUT) {
5204 					dir = PF_IN;
5205 					idx = PF_SK_STACK;
5206 				} else {
5207 					dir = PF_OUT;
5208 					idx = PF_SK_WIRE;
5209 				}
5210 
5211 				match_key.af = s->key[idx]->af;
5212 				match_key.proto = s->key[idx]->proto;
5213 				PF_ACPY(&match_key.addr[0],
5214 				    &s->key[idx]->addr[1], match_key.af);
5215 				match_key.port[0] = s->key[idx]->port[1];
5216 				PF_ACPY(&match_key.addr[1],
5217 				    &s->key[idx]->addr[0], match_key.af);
5218 				match_key.port[1] = s->key[idx]->port[0];
5219 			}
5220 
5221 			/*
5222 			 * Don't send out individual
5223 			 * delete messages.
5224 			 */
5225 			s->state_flags |= PFSTATE_NOSYNC;
5226 			pf_unlink_state(s, PF_ENTER_LOCKED);
5227 			killed++;
5228 
5229 			if (kill->psk_kill_match)
5230 				killed += pf_kill_matching_state(&match_key,
5231 				    dir);
5232 
5233 			goto relock_DIOCCLRSTATES;
5234 		}
5235 		PF_HASHROW_UNLOCK(ih);
5236 	}
5237 
5238 	if (V_pfsync_clear_states_ptr != NULL)
5239 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
5240 
5241 	return (killed);
5242 }
5243 
5244 static void
5245 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
5246 {
5247 	struct pf_kstate	*s;
5248 
5249 	if (kill->psk_pfcmp.id) {
5250 		if (kill->psk_pfcmp.creatorid == 0)
5251 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
5252 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
5253 		    kill->psk_pfcmp.creatorid))) {
5254 			pf_unlink_state(s, PF_ENTER_LOCKED);
5255 			*killed = 1;
5256 		}
5257 		return;
5258 	}
5259 
5260 	for (unsigned int i = 0; i <= pf_hashmask; i++)
5261 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
5262 
5263 	return;
5264 }
5265 
5266 static int
5267 pf_killstates_nv(struct pfioc_nv *nv)
5268 {
5269 	struct pf_kstate_kill	 kill;
5270 	nvlist_t		*nvl = NULL;
5271 	void			*nvlpacked = NULL;
5272 	int			 error = 0;
5273 	unsigned int		 killed = 0;
5274 
5275 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5276 
5277 	if (nv->len > pf_ioctl_maxcount)
5278 		ERROUT(ENOMEM);
5279 
5280 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5281 	if (nvlpacked == NULL)
5282 		ERROUT(ENOMEM);
5283 
5284 	error = copyin(nv->data, nvlpacked, nv->len);
5285 	if (error)
5286 		ERROUT(error);
5287 
5288 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5289 	if (nvl == NULL)
5290 		ERROUT(EBADMSG);
5291 
5292 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
5293 	if (error)
5294 		ERROUT(error);
5295 
5296 	pf_killstates(&kill, &killed);
5297 
5298 	free(nvlpacked, M_NVLIST);
5299 	nvlpacked = NULL;
5300 	nvlist_destroy(nvl);
5301 	nvl = nvlist_create(0);
5302 	if (nvl == NULL)
5303 		ERROUT(ENOMEM);
5304 
5305 	nvlist_add_number(nvl, "killed", killed);
5306 
5307 	nvlpacked = nvlist_pack(nvl, &nv->len);
5308 	if (nvlpacked == NULL)
5309 		ERROUT(ENOMEM);
5310 
5311 	if (nv->size == 0)
5312 		ERROUT(0);
5313 	else if (nv->size < nv->len)
5314 		ERROUT(ENOSPC);
5315 
5316 	error = copyout(nvlpacked, nv->data, nv->len);
5317 
5318 on_error:
5319 	nvlist_destroy(nvl);
5320 	free(nvlpacked, M_NVLIST);
5321 	return (error);
5322 }
5323 
5324 static int
5325 pf_clearstates_nv(struct pfioc_nv *nv)
5326 {
5327 	struct pf_kstate_kill	 kill;
5328 	nvlist_t		*nvl = NULL;
5329 	void			*nvlpacked = NULL;
5330 	int			 error = 0;
5331 	unsigned int		 killed;
5332 
5333 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5334 
5335 	if (nv->len > pf_ioctl_maxcount)
5336 		ERROUT(ENOMEM);
5337 
5338 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5339 	if (nvlpacked == NULL)
5340 		ERROUT(ENOMEM);
5341 
5342 	error = copyin(nv->data, nvlpacked, nv->len);
5343 	if (error)
5344 		ERROUT(error);
5345 
5346 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5347 	if (nvl == NULL)
5348 		ERROUT(EBADMSG);
5349 
5350 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
5351 	if (error)
5352 		ERROUT(error);
5353 
5354 	killed = pf_clear_states(&kill);
5355 
5356 	free(nvlpacked, M_NVLIST);
5357 	nvlpacked = NULL;
5358 	nvlist_destroy(nvl);
5359 	nvl = nvlist_create(0);
5360 	if (nvl == NULL)
5361 		ERROUT(ENOMEM);
5362 
5363 	nvlist_add_number(nvl, "killed", killed);
5364 
5365 	nvlpacked = nvlist_pack(nvl, &nv->len);
5366 	if (nvlpacked == NULL)
5367 		ERROUT(ENOMEM);
5368 
5369 	if (nv->size == 0)
5370 		ERROUT(0);
5371 	else if (nv->size < nv->len)
5372 		ERROUT(ENOSPC);
5373 
5374 	error = copyout(nvlpacked, nv->data, nv->len);
5375 
5376 #undef ERROUT
5377 on_error:
5378 	nvlist_destroy(nvl);
5379 	free(nvlpacked, M_NVLIST);
5380 	return (error);
5381 }
5382 
5383 static int
5384 pf_getstate(struct pfioc_nv *nv)
5385 {
5386 	nvlist_t		*nvl = NULL, *nvls;
5387 	void			*nvlpacked = NULL;
5388 	struct pf_kstate	*s = NULL;
5389 	int			 error = 0;
5390 	uint64_t		 id, creatorid;
5391 
5392 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
5393 
5394 	if (nv->len > pf_ioctl_maxcount)
5395 		ERROUT(ENOMEM);
5396 
5397 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5398 	if (nvlpacked == NULL)
5399 		ERROUT(ENOMEM);
5400 
5401 	error = copyin(nv->data, nvlpacked, nv->len);
5402 	if (error)
5403 		ERROUT(error);
5404 
5405 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
5406 	if (nvl == NULL)
5407 		ERROUT(EBADMSG);
5408 
5409 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
5410 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
5411 
5412 	s = pf_find_state_byid(id, creatorid);
5413 	if (s == NULL)
5414 		ERROUT(ENOENT);
5415 
5416 	free(nvlpacked, M_NVLIST);
5417 	nvlpacked = NULL;
5418 	nvlist_destroy(nvl);
5419 	nvl = nvlist_create(0);
5420 	if (nvl == NULL)
5421 		ERROUT(ENOMEM);
5422 
5423 	nvls = pf_state_to_nvstate(s);
5424 	if (nvls == NULL)
5425 		ERROUT(ENOMEM);
5426 
5427 	nvlist_add_nvlist(nvl, "state", nvls);
5428 	nvlist_destroy(nvls);
5429 
5430 	nvlpacked = nvlist_pack(nvl, &nv->len);
5431 	if (nvlpacked == NULL)
5432 		ERROUT(ENOMEM);
5433 
5434 	if (nv->size == 0)
5435 		ERROUT(0);
5436 	else if (nv->size < nv->len)
5437 		ERROUT(ENOSPC);
5438 
5439 	error = copyout(nvlpacked, nv->data, nv->len);
5440 
5441 #undef ERROUT
5442 errout:
5443 	if (s != NULL)
5444 		PF_STATE_UNLOCK(s);
5445 	free(nvlpacked, M_NVLIST);
5446 	nvlist_destroy(nvl);
5447 	return (error);
5448 }
5449 
5450 /*
5451  * XXX - Check for version missmatch!!!
5452  */
5453 
5454 /*
5455  * Duplicate pfctl -Fa operation to get rid of as much as we can.
5456  */
5457 static int
5458 shutdown_pf(void)
5459 {
5460 	int error = 0;
5461 	u_int32_t t[5];
5462 	char nn = '\0';
5463 
5464 	do {
5465 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
5466 		    != 0) {
5467 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
5468 			break;
5469 		}
5470 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
5471 		    != 0) {
5472 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
5473 			break;		/* XXX: rollback? */
5474 		}
5475 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
5476 		    != 0) {
5477 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
5478 			break;		/* XXX: rollback? */
5479 		}
5480 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
5481 		    != 0) {
5482 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
5483 			break;		/* XXX: rollback? */
5484 		}
5485 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
5486 		    != 0) {
5487 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
5488 			break;		/* XXX: rollback? */
5489 		}
5490 
5491 		/* XXX: these should always succeed here */
5492 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
5493 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
5494 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
5495 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
5496 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
5497 
5498 		if ((error = pf_clear_tables()) != 0)
5499 			break;
5500 
5501 #ifdef ALTQ
5502 		if ((error = pf_begin_altq(&t[0])) != 0) {
5503 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
5504 			break;
5505 		}
5506 		pf_commit_altq(t[0]);
5507 #endif
5508 
5509 		pf_clear_all_states();
5510 
5511 		pf_clear_srcnodes(NULL);
5512 
5513 		/* status does not use malloced mem so no need to cleanup */
5514 		/* fingerprints and interfaces have their own cleanup code */
5515 	} while(0);
5516 
5517 	return (error);
5518 }
5519 
5520 static pfil_return_t
5521 pf_check_return(int chk, struct mbuf **m)
5522 {
5523 
5524 	switch (chk) {
5525 	case PF_PASS:
5526 		if (*m == NULL)
5527 			return (PFIL_CONSUMED);
5528 		else
5529 			return (PFIL_PASS);
5530 		break;
5531 	default:
5532 		if (*m != NULL) {
5533 			m_freem(*m);
5534 			*m = NULL;
5535 		}
5536 		return (PFIL_DROPPED);
5537 	}
5538 }
5539 
5540 #ifdef INET
5541 static pfil_return_t
5542 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
5543     void *ruleset __unused, struct inpcb *inp)
5544 {
5545 	int chk;
5546 
5547 	chk = pf_test(PF_IN, flags, ifp, m, inp);
5548 
5549 	return (pf_check_return(chk, m));
5550 }
5551 
5552 static pfil_return_t
5553 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
5554     void *ruleset __unused,  struct inpcb *inp)
5555 {
5556 	int chk;
5557 
5558 	chk = pf_test(PF_OUT, flags, ifp, m, inp);
5559 
5560 	return (pf_check_return(chk, m));
5561 }
5562 #endif
5563 
5564 #ifdef INET6
5565 static pfil_return_t
5566 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
5567     void *ruleset __unused,  struct inpcb *inp)
5568 {
5569 	int chk;
5570 
5571 	/*
5572 	 * In case of loopback traffic IPv6 uses the real interface in
5573 	 * order to support scoped addresses. In order to support stateful
5574 	 * filtering we have change this to lo0 as it is the case in IPv4.
5575 	 */
5576 	CURVNET_SET(ifp->if_vnet);
5577 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp, m, inp);
5578 	CURVNET_RESTORE();
5579 
5580 	return (pf_check_return(chk, m));
5581 }
5582 
5583 static pfil_return_t
5584 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
5585     void *ruleset __unused,  struct inpcb *inp)
5586 {
5587 	int chk;
5588 
5589 	CURVNET_SET(ifp->if_vnet);
5590 	chk = pf_test6(PF_OUT, flags, ifp, m, inp);
5591 	CURVNET_RESTORE();
5592 
5593 	return (pf_check_return(chk, m));
5594 }
5595 #endif /* INET6 */
5596 
5597 #ifdef INET
5598 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
5599 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
5600 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
5601 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
5602 #endif
5603 #ifdef INET6
5604 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
5605 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
5606 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
5607 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
5608 #endif
5609 
5610 static void
5611 hook_pf(void)
5612 {
5613 	struct pfil_hook_args pha;
5614 	struct pfil_link_args pla;
5615 	int ret;
5616 
5617 	if (V_pf_pfil_hooked)
5618 		return;
5619 
5620 	pha.pa_version = PFIL_VERSION;
5621 	pha.pa_modname = "pf";
5622 	pha.pa_ruleset = NULL;
5623 
5624 	pla.pa_version = PFIL_VERSION;
5625 
5626 #ifdef INET
5627 	pha.pa_type = PFIL_TYPE_IP4;
5628 	pha.pa_func = pf_check_in;
5629 	pha.pa_flags = PFIL_IN;
5630 	pha.pa_rulname = "default-in";
5631 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
5632 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
5633 	pla.pa_head = V_inet_pfil_head;
5634 	pla.pa_hook = V_pf_ip4_in_hook;
5635 	ret = pfil_link(&pla);
5636 	MPASS(ret == 0);
5637 	pha.pa_func = pf_check_out;
5638 	pha.pa_flags = PFIL_OUT;
5639 	pha.pa_rulname = "default-out";
5640 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
5641 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
5642 	pla.pa_head = V_inet_pfil_head;
5643 	pla.pa_hook = V_pf_ip4_out_hook;
5644 	ret = pfil_link(&pla);
5645 	MPASS(ret == 0);
5646 #endif
5647 #ifdef INET6
5648 	pha.pa_type = PFIL_TYPE_IP6;
5649 	pha.pa_func = pf_check6_in;
5650 	pha.pa_flags = PFIL_IN;
5651 	pha.pa_rulname = "default-in6";
5652 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
5653 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
5654 	pla.pa_head = V_inet6_pfil_head;
5655 	pla.pa_hook = V_pf_ip6_in_hook;
5656 	ret = pfil_link(&pla);
5657 	MPASS(ret == 0);
5658 	pha.pa_func = pf_check6_out;
5659 	pha.pa_rulname = "default-out6";
5660 	pha.pa_flags = PFIL_OUT;
5661 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
5662 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
5663 	pla.pa_head = V_inet6_pfil_head;
5664 	pla.pa_hook = V_pf_ip6_out_hook;
5665 	ret = pfil_link(&pla);
5666 	MPASS(ret == 0);
5667 #endif
5668 
5669 	V_pf_pfil_hooked = 1;
5670 }
5671 
5672 static void
5673 dehook_pf(void)
5674 {
5675 
5676 	if (V_pf_pfil_hooked == 0)
5677 		return;
5678 
5679 #ifdef INET
5680 	pfil_remove_hook(V_pf_ip4_in_hook);
5681 	pfil_remove_hook(V_pf_ip4_out_hook);
5682 #endif
5683 #ifdef INET6
5684 	pfil_remove_hook(V_pf_ip6_in_hook);
5685 	pfil_remove_hook(V_pf_ip6_out_hook);
5686 #endif
5687 
5688 	V_pf_pfil_hooked = 0;
5689 }
5690 
5691 static void
5692 pf_load_vnet(void)
5693 {
5694 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
5695 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
5696 
5697 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
5698 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
5699 #ifdef ALTQ
5700 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
5701 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
5702 #endif
5703 
5704 	pfattach_vnet();
5705 	V_pf_vnet_active = 1;
5706 }
5707 
5708 static int
5709 pf_load(void)
5710 {
5711 	int error;
5712 
5713 	rm_init_flags(&pf_rules_lock, "pf rulesets", RM_RECURSE);
5714 	sx_init(&pf_ioctl_lock, "pf ioctl");
5715 	sx_init(&pf_end_lock, "pf end thread");
5716 
5717 	pf_mtag_initialize();
5718 
5719 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
5720 	if (pf_dev == NULL)
5721 		return (ENOMEM);
5722 
5723 	pf_end_threads = 0;
5724 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
5725 	if (error != 0)
5726 		return (error);
5727 
5728 	pfi_initialize();
5729 
5730 	return (0);
5731 }
5732 
5733 static void
5734 pf_unload_vnet(void)
5735 {
5736 	int ret;
5737 
5738 	V_pf_vnet_active = 0;
5739 	V_pf_status.running = 0;
5740 	dehook_pf();
5741 
5742 	PF_RULES_WLOCK();
5743 	pf_syncookies_cleanup();
5744 	shutdown_pf();
5745 	PF_RULES_WUNLOCK();
5746 
5747 	ret = swi_remove(V_pf_swi_cookie);
5748 	MPASS(ret == 0);
5749 	ret = intr_event_destroy(V_pf_swi_ie);
5750 	MPASS(ret == 0);
5751 
5752 	pf_unload_vnet_purge();
5753 
5754 	pf_normalize_cleanup();
5755 	PF_RULES_WLOCK();
5756 	pfi_cleanup_vnet();
5757 	PF_RULES_WUNLOCK();
5758 	pfr_cleanup();
5759 	pf_osfp_flush();
5760 	pf_cleanup();
5761 	if (IS_DEFAULT_VNET(curvnet))
5762 		pf_mtag_cleanup();
5763 
5764 	pf_cleanup_tagset(&V_pf_tags);
5765 #ifdef ALTQ
5766 	pf_cleanup_tagset(&V_pf_qids);
5767 #endif
5768 	uma_zdestroy(V_pf_tag_z);
5769 
5770 #ifdef PF_WANT_32_TO_64_COUNTER
5771 	PF_RULES_WLOCK();
5772 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
5773 
5774 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
5775 	MPASS(V_pf_allkifcount == 0);
5776 
5777 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
5778 	V_pf_allrulecount--;
5779 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
5780 
5781 	/*
5782 	 * There are known pf rule leaks when running the test suite.
5783 	 */
5784 #ifdef notyet
5785 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
5786 	MPASS(V_pf_allrulecount == 0);
5787 #endif
5788 
5789 	PF_RULES_WUNLOCK();
5790 
5791 	free(V_pf_kifmarker, PFI_MTYPE);
5792 	free(V_pf_rulemarker, M_PFRULE);
5793 #endif
5794 
5795 	/* Free counters last as we updated them during shutdown. */
5796 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
5797 	for (int i = 0; i < 2; i++) {
5798 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
5799 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
5800 	}
5801 	counter_u64_free(V_pf_default_rule.states_cur);
5802 	counter_u64_free(V_pf_default_rule.states_tot);
5803 	counter_u64_free(V_pf_default_rule.src_nodes);
5804 
5805 	for (int i = 0; i < PFRES_MAX; i++)
5806 		counter_u64_free(V_pf_status.counters[i]);
5807 	for (int i = 0; i < KLCNT_MAX; i++)
5808 		counter_u64_free(V_pf_status.lcounters[i]);
5809 	for (int i = 0; i < FCNT_MAX; i++)
5810 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
5811 	for (int i = 0; i < SCNT_MAX; i++)
5812 		counter_u64_free(V_pf_status.scounters[i]);
5813 }
5814 
5815 static void
5816 pf_unload(void)
5817 {
5818 
5819 	sx_xlock(&pf_end_lock);
5820 	pf_end_threads = 1;
5821 	while (pf_end_threads < 2) {
5822 		wakeup_one(pf_purge_thread);
5823 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
5824 	}
5825 	sx_xunlock(&pf_end_lock);
5826 
5827 	if (pf_dev != NULL)
5828 		destroy_dev(pf_dev);
5829 
5830 	pfi_cleanup();
5831 
5832 	rm_destroy(&pf_rules_lock);
5833 	sx_destroy(&pf_ioctl_lock);
5834 	sx_destroy(&pf_end_lock);
5835 }
5836 
5837 static void
5838 vnet_pf_init(void *unused __unused)
5839 {
5840 
5841 	pf_load_vnet();
5842 }
5843 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
5844     vnet_pf_init, NULL);
5845 
5846 static void
5847 vnet_pf_uninit(const void *unused __unused)
5848 {
5849 
5850 	pf_unload_vnet();
5851 }
5852 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
5853 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
5854     vnet_pf_uninit, NULL);
5855 
5856 static int
5857 pf_modevent(module_t mod, int type, void *data)
5858 {
5859 	int error = 0;
5860 
5861 	switch(type) {
5862 	case MOD_LOAD:
5863 		error = pf_load();
5864 		break;
5865 	case MOD_UNLOAD:
5866 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
5867 		 * the vnet_pf_uninit()s */
5868 		break;
5869 	default:
5870 		error = EINVAL;
5871 		break;
5872 	}
5873 
5874 	return (error);
5875 }
5876 
5877 static moduledata_t pf_mod = {
5878 	"pf",
5879 	pf_modevent,
5880 	0
5881 };
5882 
5883 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
5884 MODULE_VERSION(pf, PF_MODVER);
5885